9. Service ingress QoS policies
This section provides information to configure SAP ingress QoS policies using the command line interface.
9.1. Overview
There is one default service ingress policy. The default policy has two classification resources and one meter (the num-qos-classifiers set to value "2" and meter 1 is the default meter). The default policy uses CAM resources from the ingress-internal-tcam>qos-sap-ingress-resource pool for the classification of all traffic to the default FC, and rate-limits the traffic by using a policer that uses the default rate. SAP ingress policies with policing only are supported for SAPs configured on access ports and hybrid ports.
 | Note: Queuing and shaping on SAP ingress is not supported on the 7210 SAS-T (access-uplink and network mode), 7210 SAS-Mxp, 7210 SAS-R6, 7210 SAS-R12, 7210 SAS-Sx/S 1/10GE, and 7210 SAS-Sx 10/100GE. |
Each policy can have up to 32 ingress meters. The default policies can be copied and modified but cannot be deleted. The default policies are identified as policy ID 1.
The default policies are applied to the service entry on creation (if applicable). For example, the default SAP ingress policy is applied to access ingress SAPs. You must explicitly associate other QoS policies.
9.1.1. Default SAP ingress policy
The default policy 1 maps all traffic to default forwarding class “be” and maps FC “be” to meter “1”. Meter “1” is configured with cir 0 and pir max.
The following is a sample configuration output.
9.1.2. SAP-ingress policy defaults
The following table lists the SAP-ingress policy defaults.
Table 60: SAP-ingress policy defaults
Field | Default |
description | “Default SAP ingress QoS policy.” |
num-qos-classifiers | 2 |
scope | template |
meter | 1 |
mode | trtcm1 |
adaptation-rule | cir closest pir closest |
rate | pir = max, cir= 0 |
mbs | default |
cbs | default |
default-fc | be |
9.2. Resource allocation for SAP ingress policy
This sections describes resource allocation information for SAP ingress policies.
9.2.1. Use of index file by SAP QoS ingress policy
The 7210 SAS platforms use an index file to store the map that indicates the QoS resource allocation to SAPs. This file is used on reboot to ensure that all the SAPs that were created successfully before a reboot can be recreated after a reboot. Without an index file the system cannot ensure this (that is, without an index file it is possible that all the SAPs that were configured successfully may fail on a reboot after saving the configuration file). The file is stored in the flash memory.
On reboot, if the file is found, the system allocates resources in accordance with the stored map. If the file is not found, the system implements a best-fit algorithm and tries to allocate resources for all the SAPs on a first-come-first-served basis. If the file is not present, it is possible that the saved configuration will not execute successfully after the reboot, because the resources might not be allocated to all SAPs.
| Note: The index file used for the QoS map is different from the one used for storing interface indexes. |
9.2.2. Use of the keyword “multipoint” for default meter “11”
The system allows sharing of a single meter for both unicast and multipoint traffic. Users can configure any of the available meters for multipoint traffic. The use of the multipoint keyword during meter creation has been deprecated, except for use with meter “11”, as described in the following paragraphs.
When the multipoint keyword is specified with meter "11" the system interprets it to be the default multipoint meter. The default multipoint meter is used for all FCs that do not have explicit multipoint meters configured. The system does the appropriate resource checks to ensure that resources needed to use multipoint meter with all the FCs are available before allowing this change.
| Note:
|
9.2.2.1. Example uses of the multipoint meter
This section provides configuration examples of several uses of the multipoint meter.
Example 1
In example 1, all FCs in the SAP-ingress policy use the default meter 1 (for all traffic types). If the configure qos sap-ingress id meter 11 multipoint create command is executed, it attaches the default meter 11 with all the FCs defined in the SAP-ingress policy.
After configuring meter 11 multipoint, all the FCs in this policy use two meters: default meter 1, to meter unicast traffic for all the FCs; and meter 11, to meter BUM traffic for all the FCs. In this example, because only the default FC “be” is in use, the multipoint meter is used to meter BUM traffic associated with default FC “be”.
The following example shows the policy after the configuration is changed.
Deleting the multipoint meter 11 removes all the FCs associated with the multicast-meter, assuming all the FCs are using the default multicast meter and do not have any other multicast meter explicitly configured. Executing the configure qos sap-ingress id no meter 11 command disassociates meter 11 from the FCs, and the FCs use only meter 1 (if no other meter is configured explicitly).
Example 2
Starting with the policy in example 2, if the user now executes the configure qos sap-ingress id meter 11 multipoint create command, the FC “be” continues to use meter 3 and the FC “af” uses meter 11 for BUM traffic. In this example, if the user executes the configure qos sap-ingress id fc be no multicast-meter command, the default meter 11 is also used for FC “be”.
Example 3
Upon executing the configure qos sap-ingress id meter 11 multipoint create command, FC "be" unknown-unicast traffic type continues to use meter 3, and broadcast and multicast traffic types use meter 11.
In example3, if a broadcast-meter was initially configured in the SAP-ingress policy and was followed by executing the configure qos sap-ingress id meter 11 multipoint create command, FC “be” changes to use meter 11 for multicast traffic, and broadcast traffic continues to use meter 3 for unknown-unicast traffic and meter 3 for unicast traffic.
Also in example 3, if the user executes the configure qos sap-ingress id fc be no unknown-meter command, meter 3 is used for all traffic types classified to FC “be”. But, if the default meter 11 is defined in the policy, FC “be” uses meter 11 for BUM traffic.
9.2.3. Service ingress meter selection rules
The following are rules for meter selection by different traffic types under various configurations.
9.2.3.1. Default policy
In the default policy, only meter “1” is defined. All FC and all traffic types use meter “1” by default. Meter “11” is not created by default and is not available for use.
The following is a sample configuration output.
9.2.3.2. VPLS service without meter “11”
The following describes the usage of meters when meter “11” in a VPLS service is not configured in the policy:
- If an FC is created without explicit meters, the default meter “1” is used for unicast traffic and for multipoint traffic types (such as broadcast, multicast and unknown-unicast traffic).
- If an FC is created with an explicit unicast meter, that meter is used for unicast traffic and for multipoint traffic types (such as broadcast, multicast and unknown-unicast traffic).
- If an FC is created with an explicit unicast meter and explicit broadcast meter, use these meters for unicast and broadcast traffic respectively and use the unicast meter for all other traffic types.
- If an FC is created with an explicit unicast meter and explicit multicast meter, use the unicast meter for unicast traffic and multicast meter for all other traffic types.
- If an FC is created with an explicit unicast meter, an explicit broadcast meter, and an explicit multicast meter, use these meters for unicast, broadcast and multicast traffic types respectively. Unknown unicast traffic type will use the explicitly defined multicast meter.
- If an FC is created with an explicit unicast meter, an explicit broadcast meter, an explicit unknown-unicast meter, and an explicit multicast meter, use these meters for unicast, broadcast, unknown-unicast and multicast traffic types respectively.
9.2.3.3. VPLS service with meter “11”
The following describes the usage of meters when meter “11” in a VPLS service is defined in the policy:
- If an FC is created without explicit meters, use the default meter “1” for unicast traffic and default meter “11” for all other traffic types (such as broadcast, multicast and unknown-unicast).
- If an FC is created with an explicit unicast meter, use that meter for unicast traffic and use default meter “11” for all other traffic types.
- If an FC is created with an explicit unicast meter and explicit broadcast meter, use these meters for unicast and broadcast traffic respectively and use meter “11” for all other traffic types.
- If an FC is created with an explicit unicast meter and explicit multicast meter, use the unicast meter for unicast traffic and multicast meter for all other kinds of traffic.
- If an FC is created with an explicit unicast meter, an explicit broadcast meter, and an explicit multicast meter, use these meters for unicast, broadcast and multicast traffic types respectively. Unknown unicast traffic type will use the explicitly defined multicast meter.
- If an FC is created with an explicit unicast meter, an explicit broadcast meter, an explicit unknown-unicast meter, and an explicit multicast meter, use these meters for unicast, broadcast, unknown-unicast and multicast traffic types respectively.
9.2.3.4. Epipe, IES, and VPRN services without PIM
The following are rules for meter selection for Epipe, IES and VPRN services:
- A multipoint meter cannot be used. A multipoint meter configured in a policy is not used when the policy is applied to a SAP in an Epipe service.
- All FCs associated with a meter always use the unicast meter.
| Note: These rules apply to IES services when PIM is not enabled in the service. |
9.2.3.5. IES and VPRN services with PIM/multicast and without meter “11”
The following are rules for meter selection for IES and VPRN services when PIM/multicast is enabled in the service and describes the usage of meters when meter “11” is not configured in the policy:
- If an FC is created without explicit meters, the default meter “1” is used for unicast traffic and multicast traffic.
- If an FC is created with an explicit unicast meter, that meter is used for unicast traffic and for multicast traffic.
- If an FC is created with an explicit unicast meter and explicit multicast meter, use the unicast meter for unicast traffic and multicast meter for multicast traffic.
9.2.3.6. IES and VPRN services with PIM/multicast and meter “11”
The following are rules for meter selection for IES and VPRN services when PIM/multicast is enabled in the service and when meter “11” is defined in the policy:
- If an FC is created without explicit meters, use the default meter “1” for unicast traffic and default meter “11” for multicast traffic.
- If an FC is created with an explicit unicast meter, use that meter for unicast traffic and use default meter “11” for multicast traffic.
- If an FC is created with an explicit unicast meter and explicit multicast meter, use the unicast meter for unicast traffic and multicast meter for multicast traffic.
9.2.4. Service ingress policy configuration considerations
The num-qos-classifiers parameter cannot be modified when the policy is in use (for example, when it is associated with a SAP). Other parameters in the SAP ingress policy can be changed.
When changing other parameters for a policy that is in use (for example, fc meter map or fc classification match criteria entries), the system recomputes the resources required to accommodate the change. If the resources required exceeds the configured value for num-qos-classifiers, the change is not allowed.
If more resources are needed than the value configured in num-qos-classifiers for a existing policy, then the following options are available:
- Copy the existing policy to a new policy, modify the num-qos-classifiers parameter, modify the match criteria entries, and modify the SAP configuration to associate it with the new policy.
- Ensure the existing policy is not in use by any SAP (if required, change the SAP configuration to disable the use of the QoS policy with the no qos form of this command), change all the required parameters, and finally modify the SAP configuration to use the policy again.
Note: Both options above have side effects; for example, they can reset the statistics associated with the meters and can potentially cause existing traffic classification not to take effect. However, the system ensures that the default policy is used during the period when policy changes are being made after the two options are performed.
The following items are additional service ingress policy configuration considerations:
- In releases prior to Release 3.0R1, the system always the computes the number of resources (like classifiers and meters) required by a policy assuming the number of resources will be used in a VPLS service. This allows the policy to be applied to either an Epipe or VPLS service.
- From Release 3.0R1 onwards, on creation of SAP ingress policy, the system does not compute the number of resources required by a policy and validate the number against resources available in the system. The system validates the resources needed only when the SAP ingress policy is attached to a SAP. If enough resources are available the association succeeds, otherwise the system fails the CLI command. Based on the service in which the SAP is configured (such as VLL, VPLS, and so on), for the same SAP ingress policy the amount of resources required is different. The system validates that the amount of QoS resources specified in the num-qos-classifiers command is sufficient for the match criteria, forwarding class, and service specified and that the resources are available in hardware. On failure of the validation, the system disallows the association of the SAP ingress policy with the SAP.
- The match criteria type (that is, mac-, ipv4-, and ipv6-criteria) cannot be changed when the SAP ingress QoS policy is in use. For example, if the match-criteria is set to ipv4-criteria and the policy is associated with a SAP, then the ipv6-criteria or mac-criteria cannot be enabled in the same policy. If there is a need to change the criteria, users must remove the association and then change the SAP ingress policy to use the new match criteria. For SAPs configured in VPRN services, the computation of resources is similar to an SAP configured in an Epipe service.
See Resource allocation for service ingress QoS policies using CAM-based classification for more information.
9.2.5. Resource allocation for service ingress QoS policies using CAM-based classification
The available global pool of ingress internal CAM hardware resources can be allocated based on user needs and shared among different features, such as SAP ingress QoS policy, ingress ACLs, and so on. Use the configure>system>resource-profile>ingress-internal-tcam>qos-sap-ingress-resource command to allocate SAP ingress QoS classification and meter resources from this global pool.
In addition, resources can be allocated for different SAP ingress QoS policy classification match criteria based on operator needs. Users can modify resource allocation to scale the number of entries available per match criteria or to scale the number of SAPs. Resources from the global ingress internal CAM pool are allocated in fixed slices. The number of classification entries and meters per slice varies across 7210 SAS platforms.
For more information about number of slices that can be allocated to a feature, refer to the System Resource Allocation section in the 7210 SAS-Mxp, R6, R12, S, Sx, T Basic System Configuration Guide. For example, on a 7210 SAS-R6 IMM-b card, each slice allows 256 classification entries and 128 meters. The number of slices allotted for a SAP ingress QoS policy is configured using the configure>system>resource-profile>ingress-internal-tcam>qos-sap-ingress-resource command.
Users can configure the resources available for SAP ingress QoS policies and can limit the amount of resources used per match criteria supported for SAP ingress QoS policies. A specific slice can be used for either MAC criteria or IP criteria or IPv6 criteria, or both MAC and IP criteria. Allocation of classification entries also allocates meter/policer resources that are used to implement per FC per traffic type policing.
| Note: For SAP ingress classification, in addition to CAM-based resource allocation, the 7210 SAS-Mxp, 7210 SAS-R6, and 7210 SAS-R12 support table-based resource allocation for DSCP-classification on SAP ingress. See Table-based classification using dot1p and IP DSCP for assigning FC and profile on SAP ingress for the 7210 SAS-Mxp, 7210 SAS-R6, and 7210 SAS-R12 for more information. |
By default, the system allocates resources for SAP ingress QoS policies to maintain backward compatibility with Release 4.0 and allocates resources for MAC criteria and IP criteria (by setting num-resources to max). Setting the num-resources value to max allows each match criteria to use the available SAP ingress QoS resources on a first-come-first-served model. By default, system does not allocate resources for use by ingress IPv6 classification or for using both IP (any) and MAC (any) criteria in a single SAP ingress policy. You must allocate resources before associating a SAP with an IPv6 SAP ingress policy or both IP (any) and MAC (any) criteria in a SAP ingress policy. Until appropriate resources are allocated, attempts to associate the policy with the SAP will fail.
If the configure>system>resource-profile>qos-sap-ingress-resource command is used to allocate resources for SAP ingress QoS policies, the system allocates resources in slices for a fixed number of entries; the entries allocated per slice is platform-dependent. The use of these entries by different types of match criteria is described in the following table.
Table 61: SAP ingress resource allocation and match criteria types
Type of match criteria | Description |
mac-criteria (any) | Before using SAP-ingress policies with mac-criteria (any), use the configure system resource-profile ingress-internal-tcam qos-sap-ingress-resource mac-match-enable CLI command to allocate resources from the SAP-ingress QoS resource pool. Resources are allocated for SAP-ingress policies that use mac-criteria (any and dot1p-only). Each entry in the SAP-ingress QoS policy that is configured to use mac-criteria uses one entry from the slices in the hardware resource pool allocated to mac-criteria. For example: Assume a SAP ingress QoS policy is configured to use mac-criteria with 50 entries and uses the configure> system> resource-profile> ingress-internal-tcam> qos-sap-ingress-resource> mac-match-enable 1 CLI command to configure one slice for use by mac-criteria (allowing a total of 512 entries for use by policies that use mac-criteria). In this case, the user can configure 10 SAPs that use the mac-criteria SAP ingress policy and consume 500 entries. |
ipv4-criteria (any) | The use of ipv4-criteria (any) match criteria is the same as the mac-criteria. Use the configure system resource-profile ingress-internal-tcam qos-sap-ingress-resource ipv4-match-enable CLI command to allocate resources. Additionally, IPv4 criteria can share entries allocated for IPv6 criteria. The SR OS automatically allocates entries from an IPv6 criteria slice to IPv4 criteria policies, if no entries are available in the allocated IPv4 criteria slices and no slices are available for allocation to IPv4 criteria from the SAP-ingress QoS resource pool. If an IPv4 criteria entry uses IPv6 criteria slices, the number of hardware entries used is the same as required by an IPv6 criteria entry (see ipv6-criteria (any) for more information). |
ipv6-criteria (any) | Before using the ipv6-criteria match criteria, use the configure> system> resource-profile> ingress-internal-tcam> qos-sap-ingress-resource> ipv6-ipv4-match-enable CLI command, and specify the ipv6 keyword for the num-qos-classifiers command to allocate resources from the SAP ingress QoS resource pool. Each ipv4-criteria match entry or ipv6-criteria configured in the QoS policy that uses ipv6-criteria uses two (2) entries from the slices allocated for use by ipv6-criteria (128-bit) in the hardware. The system allocates entries from the ipv6-criteria pool in the following cases:
For example: Assume a QoS policy is configured to use ipv6-criteria with 50 entries and the configure>system> resource-profile> ingress-internal-tcam> qos-sap-ingress-resource> ipv4-ipv6-128-match-enable 1 command is used to configure one slice for ipv6-criteria. This allows a total of 256 entries for use by SAPs using SAP ingress QoS policies with ipv6-criteria (because each IPv6 entry uses 2 entries in hardware). In this example, the user can configure five (5) SAPs that use this policy and consume a total of 250 entries. These resources can be shared with policies that use IPv4 criteria, though each IPv4 criteria entry consumes two (2) entries in the hardware. IPv4 criteria policies can consume spare IPv6 resources; however, if a larger number of IPv4 criteria policies are planned, it is good practice to allocate more resources for use with IPv4 criteria. Note: When a slice is allocated for IPv6 criteria, the system automatically adjusts the number of available classification entries in that slice to half the number of total entries available. For example on 7210 SAS-T, if a slice with a total of 256 entries is allocated for IPv6 criteria, the actual number of classification entries available for IPv6 classification is only 128. The number of meters available is 128 in this example. |
IPv4 (any) and MAC (any) match | Before using IP-criteria (any) and MAC-criteria (any) in a single policy, use the configure> system> resource-profile> ingress-internal-tcam> qos-sap-ingress-resource> ipv4-mac-match-enable command to allocate resources from the SAP ingress QoS resource pool. Each ipv4-criteria match entry or MAC-criteria match entry configured in the QoS policy uses two (2) entries from the allocated slices. The system allocates entries from the ipv4-mac-match-enable pool if the SAP ingress QoS policy uses both ip-criteria (any) and ipv6-criteria (any). The system also allocates entries for all other criteria if there are no resources available to use in the pool allocated to those criteria. That is, if no resources are available in other pools, the following criteria are allocated resources from this pool: only mac-criteria any, only ip-criteria any, mac-criteria dot1p-only, ip-criteria dscp-only, ipv4-criteria dscp-only. Note: Resources are not allocated to ipv6-criteria (any) from this pool. |
dot1p-only, IPv4 dscp-only, IPv6 dscp-only, and default SAP ingress QoS policies | Use the dot1p-only or dscp-only option if only dot1p bits or only IP DSCP bits or only IP precedence bits are used for SAP ingress classification. This facilitates efficient use of available hardware resources and better scaling. SAP ingress policies that use only dot1p bits or only IPv4/IPv6 DSCP or IPv4/IPv6 precedence or default SAP ingress QoS policy bits can use the resources from slices currently allocated for use by either IP-criteria or MAC-criteria or IPv6 criteria. The following is a list of special cases for resource allocation for default, dot1p-only, and dscp-only SAP ingress policies are as follows:
|
The SAP ingress resource slices described in this section are different from the resources that are configured using the num-qos-classifiers command.The num-qos-classifiers command sets the limit on the resources needed per SAP ingress QoS policy. The qos-sap-ingress-resource resources set the maximum limit on the resources available to all the SAP ingress policies that are in use simultaneously on the system.
The SR OS manages the resource slices allocated to the SAP ingress QoS policy pool and allocates the slice entries when a SAP ingress QoS policy is associated with a SAP. In other words, a SAP specifies the number of QoS resources it needs using the num-qos-classifiers command (in the SAP ingress policy), while the system allocates the resources required by the SAP from the qos-sap-ingress-resource slices, depending on whether the SAP ingress policy uses ip-criteria or mac-criteria or ipv6-criteria.
Use the tools dump system-resources command to view the current usage and availability of system resources. One or more entries per slice are reserved for system use.
9.2.5.1. Resource configuration guidelines for service ingress QoS policies using CAM-based classification
When a SAP is associated with a default SAP ingress QoS policy and there are no resources available in the pool of slices already allocated for different criteria that are in use, a new slice is allocated and set to either mac-match criteria, ipv4-match criteria, or ip-dscp-port-if-match criteria. This allocation can result in a single slice getting consumed and becoming unavailable for other classification criteria even if the mac-match criteria, ipv4-match criteria, or ip-dscp-port-if-match criteria are not used. To prevent this scenario, the SAP ingress resource configuration can be set to the specific number of slices for each criteria in use so that the SR OS can allocate the slices based on user requirement without allocating resources for any of the mac-match criteria, ipv4-match criteria, or ip-dscp-port-if-match criteria.
9.2.6. Computation of resources used per SAP ingress policy for CAM-based classification
Users can configure the number of classification entries the SAP requires (TQ). The value of TQ is set using the num-qos-classifiers command, where TQ is the total number of QoS resources required by the SAP. To determine TQ, see Determining the number of policers/meters per policy (TP).
Number of meters allocated automatically by system = TQ / 2 (up to a maximum of 32 meters)
To calculate the number of SAPs allowed, assume all SAPs are configured to use “TQ” QoS resources per SAP.
Number of SAPs allowed = maximum classification entries / TQ
| Note: The number of SAPs calculated using the equation above is subject to system limits. The above equation is used to derive the limit on the number of SAPs due to QoS resources only. |
Users can mix and match SAPs with different QoS resources (that is, use different values of TQ).
The following criteria determine the number of QoS resources allocated for a SAP:
- number of match-criteria entries used to identify the FC
- number of FCs to use and number of traffic types to be policed per FC
- amount of hardware classification resources needed per entry configured by the user; see Resource allocation for service ingress QoS policies using CAM-based classification for more information about resources needed per match entry, which varies based on the different match criteria in use
Only FCs that are in use by the match-criteria classification entries are considered for the number of FCs. These FCs are referred to as “FC in use”.
9.2.6.1. Determining the number of classification entries
This section describes the rules and methods of determining the number of classification entries.
9.2.6.1.1. Rules for a SAP in a VPLS
Knowing the number of traffic types to use per “FC in use”, apply the following rules for a SAP in a VPLS service to determine the number of classification entries per “FC in use”:
- If an FC is in use and is created without explicit meters, use default meter “1” for unicast traffic and for all other traffic types (that is, broadcast, multicast and unknown-unicast). This requires one classification entry in hardware. This assumes default mulitpoint meter “11” is not created by the user.
- If an FC is in use and is created without explicit meters, use default meter “1” for unicast traffic and default meter “11” (assuming meter “11” is created by the user), for all other traffic types (that is, broadcast, multicast and unknown-unicast). This requires two classification entries in hardware.
- If an FC is in use and is created with an explicit unicast meter, use that meter for unicast traffic and for all other traffic types (that is, broadcast, multicast and unknown-unicast). This requires one classification entries in hardware. This assumes default multipoint meter “11” is not created by the user.
- If an FC is in use and is created with an explicit unicast meter, use that meter for unicast traffic and use default meter “11” (assuming meter “11” is created by the user) for all other traffic types. This requires two classification entries in hardware.
- If an FC is in use and is created with an explicit unicast meter and explicit broadcast meter, use these meters for unicast and broadcast traffic respectively and use the unicast meter for all other traffic types (that is, multicast and unknown-unicast). This requires two classification entries in hardware. This assumes that the default multipoint meter “11” is not created by the user.
- If an FC is in use and is created with an explicit unicast meter and explicit broadcast meter, use these meters for unicast and broadcast traffic respectively and use meter “11” (assuming meter “11” is created by the user) for all other traffic types. This requires three classification entries in hardware.
- If an FC is in use and is created with an explicit unicast meter and explicit multicast meter, use the unicast meter for unicast traffic and multicast meter for all other kinds of traffic. This requires two classification entries in hardware.
- If an FC is in use and is created with an explicit unicast meter, an explicit broadcast meter, and an explicit multicast meter, use these meters for unicast, broadcast and multicast traffic types respectively. Unknown unicast traffic type will use the explicitly defined multicast meter. This requires three classification entries in hardware.
9.2.6.1.2. Rules for a SAP in a VLL, VPRN, or IES service with PIM disabled
Apply the following rules for a SAP in a VLL, VPRN, or IES service with PIM disabled, to determine the number of classification entries per FC:
- Multipoint meters cannot be used. Multipoint meter configured in a policy is not used when the policy is applied to a SAP in an Epipe service.
- All FCs in use and associated with a meter always use the unicast meter. Therefore, all FCs in use utilize only one classification entry in the hardware.
9.2.6.1.3. Rules for a SAP in an IES or VPRN service with PIM/multicast enabled
Knowing the number of traffic types to use per “FC in use”, apply the following rules for a SAP in an IES or VPRN service enabled with PIM/multicast enabled to determine the number of classification entries per FC in use:
- If an FC is in use and is created without explicit meters, use default meter “1” for unicast traffic and for multicast traffic. This requires one classification entry in hardware. This assumes default multipoint meter “11” is not created by the user.
- If an FC is in use and is created without explicit meters, use default meter “1” for unicast traffic and default meter “11” (assuming meter “11” is created by the user), for multicast traffic. This requires two classification entries in hardware.
- If an FC is in use and is created with an explicit unicast meter, use that meter for unicast traffic and for multicast traffic. This requires one classification entries in hardware. This assumes default multipoint meter “11”” is not created by the user.
- If an FC is in use and is created with an explicit unicast meter, use that meter for unicast traffic and use default meter “11” (assuming meter “11” is created by the user) for multicast traffic. This requires two classification entries in hardware.
- If an FC is in use and is created with an explicit unicast meter and explicit multicast meter, use the unicast meter for unicast traffic and multicast meter for multicast traffic. This requires two classification entries in hardware.
9.2.6.1.4. Calculating the number of classification entries per FC
Apply the rules (above) to determine the number of classification entries per FC (Ci) (only for the “FCs in use”) using the following equation:
C(i) = ∑FCi (unicast) + FCi (multicast) + FCi (broadcast) + FCi (unknown_unicast)
i = nc, h1, ef, h2, l1, af, l2, be
where:
- FCi (unicast), FCi (multicast), FCi (broadcast), and FCi (unknown-unicast) are set to a value of 1 if the FC uses a classifier to identify traffic-type unicast, multicast, broadcast, and unknown-unicast, respectively
- FCi (unicast), FCi (multicast), FCi (broadcast), and FCi (unknown-unicast) are set to a value of 0 if the FC does not use a classifier to identify the traffic-type
If the user does not configure meters explicitly for the FC and multipoint meter “11” is not created, the default unicast meter is used for all traffic types and therefore, only one classification entry in hardware is required by the FC. If the user does not configure meters explicitly for the FC and multicast meter “11” is created, the default unicast meter and the default multicast meter are used. Therefore, by default, two classification entries in hardware are required by an FC.
9.2.6.1.5. Determining the number of classification entries per policy (TC)
Taking into account the number of match criteria and the number of FCs used, use the equation below to determine the total number of classification entries per policy, for example:
TC = ∑ E(i) * C(i)
i = nc, h1, ef, h2, l1, af, l2, be
where:
- E(i) is the number of match-criteria entries that classify packets to FCi. For 7210 SAS platforms, the maximum number of classification entries per policy can be 64 (including default).
- C(i) is the number of classification entries that are required by FCi to identify different traffic types.
9.2.6.1.6. Determining the number of policers/meters per policy (TP)
Determine the number of policers or meters to use (TP). A maximum of 32 meters per policy are available. The number of policers/meters is determined by the number of meters associated with FCs in the SAP-ingress QoS policy.
Use the values of TC and TP to determine the required number of QoS resources (TQ).
Only those meters associated with FCs are considered for the number of meters. Note that only “FCs in use” are considered.
Total QoS resources required (TQ) = max [(TC), (2 * TP))
The resulting number is rounded up to the next multiple of “2” greater than TQ obtained above. For example, if TC = 5 and TP = 2, then max (5, (2 * 2)) is 5, and TQ is rounded up to 6.
The user configures TQ value using num-qos-classifiers command.
For more information and examples about resource calculation, see the following sections:
9.3. Table-based classification using dot1p and IP DSCP for assigning FC and profile on SAP ingress for the 7210 SAS-Mxp, 7210 SAS-R6, and 7210 SAS-R12
Using an IP DSCP classification policy and a dot1p classification policy is another method used to assign the FC and profile on SAP ingress for use with color-aware meters. Similar to network ingress, you can use an IP DSCP classification policy and a dot1p classification policy to assign the FC and profile on SAP ingress for use with color-aware meters. In this section, this classification is also called table-based classification. Table-based classification is supported on the 7210 SAS-Mxp both in low-sap-scale mode and high-sap-scale mode. Table-based classification is mutually exclusive with CAM-based classification.
This is supported along with the capability to use other header fields in MAC and IP packets through the use of the mac-criteria, ipv4-criteria, and ipv6-criteria commands. When using SAP-ingress color-aware meters and policers, users can configure DEI to assign the initial profile on ingress and can configure either MAC criteria or IP criteria (or both) to assign the FC.
To configure IP DSCP classification, users create a DSCP classification policy, associate the policy with a SAP ingress QoS policy or an Ethernet ingress port or L3 interface (as applicable), apply the SAP ingress QoS policy to the SAP or port, and enable the use of the policy.
Similarly, to configure dot1p classification, users create a dot1p classification policy, associate the policy with a SAP ingress QoS policy or an Ethernet ingress port, apply the SAP ingress QoS policy to the SAP or port, and enable the use of the policy.
DSCP and dot1p classification use classification resources from a table classifier and, potentially, a lesser number of resources from the CAM resources, thereby saving CAM resources for other purposes. When a table-based classification policy is enabled, CAM-based classifications from the SAP ingress QoS policy are ignored.
The following topics describe the support for FC and profile assignment based on IP DSCP and dot1p on SAP ingress.
9.3.1. IP DSCP and dot1p classification policy support
This section describes IP DSCP and dot1p classification support.
For SAPs in Layer 2 services (Epipe and VPLS) configured to use table-based classification, you can use the config>qos>sap-ingress>table-classification-criteria CLI command, which provides the option to select one of the following: dot1p or IP DSCP, or both dot1p and IP DSCP, or none. This command is applicable only to SAPs configured in Layer 2 services; it is ignored for Layer 3 services (IES, VPRN and RVPLS services).
The following behavior is supported with table-based classification for SAPs configured in Layer 2 services depending on the configuration option selected with the config>qos>sap-ingress>table-classification-criteria CLI command:
- If none is configured, use default-fc fc-name profile out (from the SAP ingress policy).
- If use-dscp is configured, use the following policies:
- DSCP classification policy for IP packets
- default-fc fc-name profile out (from the SAP ingress policy) for non-IP packets
- If use-dot1p is configured, use the following policies:
- dot1p classification policy for all tagged packets (IP and non-IP)
- default-fc fc-name profile out (from the SAP ingress policy) for untagged packets
- If both-dscp-dot1p is configured, use the following policies:
- DSCP classification policy for IP packets
- dot1p classification policy for non-IP tagged packets
- default-fc fc-name profile out (from the SAP ingress policy) for non-IP untagged traffic
Table-based classification for SAPs configured in Layer 3 services (IES and VPRN) does not allow classification options with the config>qos>sap-ingress>table-classification-criteria CLI command. It must always be set to both-dscp-dot1p. The following behavior is supported:
- IP DSCP-based classification to assign FC and profile to an IP packet (both IPv4 and IPv6) received on SAP ingress
| Note: All SAPs in an RVPLS service can use either table-based classification or CAM-based classification entries. There cannot be a mix of SAPs where some SAPs use table-based classification and others use CAM entries. |
The following is the order of match for packets with table-based classification in Layer-3 services:
- IP packets are matched against IP DSCP entries. Ideally, packets match one of the explicitly configured user entries that classifies packets to the configured FC and profile. If there is no match, packets are assigned a default-dscp-fc and profile. If a packet is a non-IP packet, go to 2.
- Non-IP tagged Ethernet packets that do not match any IP DSCP entries are matched against dot1p values. Ideally, packets match one of the explicitly configured user entries that classifies packets to the configured FC and profile. If there is no match, packets are assigned a default-dot1p-fc and profile. Alternatively, users have an option to use Drop Eligible Indicator (DEI) to assign the profile. If a packet is an untagged non-IP packet, go to item 3.
- All non-IP untagged Ethernet packets are assigned a default-fc/untagged-fc (user-configurable) value for FC and profile out.
9.3.1.1. Default-FC assignment rules for SAPs in Layer 3 services
The following default-fc assignment rules apply for SAPs configured in Layer 3 services, including RVPLS services:
- Bridged IP packets processed in an RVPLS context that do not match any of the explicitly configured DSCP classification entries in the access port DSCP classification policy are assigned the default-dscp-fc configured in the policy. The profile assigned on ingress is in accordance with the user configuration (for example, default-dscp-fc "be" profile out).
- Bridged, tagged non-IP Layer 2 Ethernet packets, which are processed in an RVPLS context and that do not match any of the explicitly configured dot1p classification entries in the access port dot1p classification policy, are assigned the default-dot1p-fc configured in the policy. The profile assigned on ingress is in accordance with the user configuration (for example, default-dot1p-fc "be" profile out).Bridged, untagged non-IP Layer 2 Ethernet packets, which are processed in an RVPLS context, will not match any of the explicitly configured dot1p classification entries in the access port dot1p classification policy. They are assigned access port untagged-fc profile.
- Routed IP packets processed in an RVPLS context that do not match any of the explicitly configured DSCP classification entries in the DSCP classification policy associated with the interface using the routed-override-qos-policy command are assigned the default-dscp-fc configured in the policy. The profile assigned on ingress is in accordance with user configuration in the DSCP classification policy.
- IP packets processed in an IES or VPRN service context that do not match any of the explicitly configured DSCP classification entries in the DSCP classification policy associated with SAP ingress policy are assigned the default-dscp-fc configured in the DSCP classification policy. The profile assigned on ingress is in accordance with user configuration in the DSCP classification policy.
9.3.2. Precedence rules for DEI assignments on the 7210 SAS-Mxp, 7210 SAS-R6, and 7210 SAS-R12
The following precedence rules pertain to the use of DSCP/dot1p classification policies and DEI profile assignment:
- Assigning a profile using an IP DSCP/dot1p classification policy has precedence over assigning a profile using DEI bits. That is, if both a DEI-based initial profile assignment and a DSCP-based initial profile assignment are enabled on a port, the profile assigned by DSCP/dot1p-classification policy takes precedence for IP and non-IP packets, respectively.
- The CLI command to enable DEI-based profile assignment (configure>port> ethernet>enable-dei) can be used to assign a profile for SAPs configured to use CAM-based classification entries:
- A port can have a mix of SAPs with some SAPs configured to use DEI for profile assignment with CAM-based classification and other SAPs configured to use table-based classification for both FC and profile assignment.
- The enable-dei command does not affect SAPs that use table-based classification. That is, a profile assignment for SAPs configured to use table-based classification always uses a DSCP or dot1p classification policy.
9.3.3. Creating an IP DSCP and dot1p classification policy
IP DSCP and dot1p classification policies can be used at SAP ingress, allowing users to define the mapping of an IP DSCP or dot1p value to an FC and profile.
The default values for the default-dscp-fc and default-dot1p-fc command are FC “be” and profile “out”. Newly-created classification policies contain the default “be” and “out” values as the default entries. The default-dscp-fc or default-dot1p-fc command assigns the default FC to any IP DSCP or dot1p value that is not explicitly configured by a user.
Up to 50 unique DSCP or dot1p classification policies can be supported on the 7210 SAS-Mxp, 7210 SAS-R6, and 7210 SAS-R12.
Use the following syntax to create a DSCP classification QoS policy.
The following example shows the command usage to create a DSCP classification QoS policy.
Use the following syntax to create a dot1p classification QoS policy.
The following example shows the command usage to create a dot1p classification policy.
9.3.4. CAM resource usage for IP DSCP and dot1p classification policies on the 7210 SAS-Mxp, 7210 SAS-R6, and 7210 SAS-R12
The resources for IP DSCP and dot1p classification policies are taken from hardware tables, which is referred to as table-based classification to differentiate it from CAM-based classification. Table-based resources do not use many CAM entries for classification. Only a fixed number of CAM resources is needed to match the FC and traffic type and to assign a meter/policer. Table-based classification uses CAM resources more efficiently than CAM-based classification:
- Users can enable an IP DSCP and dot1p classification policy on a per-SAP or per-port basis. See Associating a DSCP or dot1p classification policy.
- The rules for associating an IP DSCP and dot1p classification policy with SAPs for different services are different. See Assigning and enabling DSCP and dot1p classification policies to a SAP.
To calculate the number of resources needed, see the following sections:
9.3.4.1. CAM resource allocation for table-based classification
When table-based classification is used from the SAP ingress resource pool, CAM resources are used to match the FC and the traffic type (unicast, broadcast, multicast, and unknown) and to assign a meter to the FC. The maximum number of DSCP entries required is 36 entries [8 FCs * 4 traffic types + 1 default FC * 4 traffic types). The maximum number of CAM resources required is 32, which assumes the use of VPLS service with one meter per traffic type, regardless of the number of IP DSCP classification entries. In other words, even if 64 IP DSCP values are matched, only 32 classification resources in the SAP ingress CAM resource pool are needed. For Epipe services, the number of entries is reduced to nine (8 FCs + 1 default FC) since all traffic is treated as unicast.
To allocate resources to meters for a SAP ingress QoS policy that is using table-based DSCP classification, use the ip-dscp-port-if-match-enable num-resources command, found under the configure> system> resource-profile> ingress-internal-tcam> qos-sap-ingress-resource context. The command supports up to 10 resources (meters).
Use the following syntax to allocate resources to meters for table-based classification.
The following example shows the command usage to allocate resources to meters for table-based classification.
9.3.5. Associating a DSCP or dot1p classification policy
A DSCP or dot1p classification policy must be associated with a SAP ingress QoS policy or an Ethernet port to map flows to an FC and profile for IP traffic received on SAP ingress.
The dscp-classification policy-id or dot1p-classification policy-id value identifies which classification policy is used to match IP packets and map the IP DSCP or dot1p to an FC and profile. The no form of the dscp-classification or dot1p-classification command associates the default classification policy (policy 1). The default policy maps all IP DSCP or dot1p values to FC “be” and profile “out”. If a packet does not match any explicitly configured criteria in the policy, the default-fc mapping is used. For details about Layer 2 and Layer 3 scenarios, see IP DSCP and dot1p classification policy support.
9.3.5.1. Associating a classification policy with a SAP ingress QoS policy
The num-qos-classifiers command allocates meters from the resources allocated towards the qos-sap-ingress pool, from the ingress-internal-tcam resource pool.
MAC, IPv4, and IPv6 criteria do not need to be defined because DSCP classification or dot1p classification is being used.
Use the following syntax to associate an IP DSCP classification policy with a SAP ingress policy.
The following example shows the command usage to associate an IP DSCP classification policy with a SAP ingress policy.
Use the following syntax to associate a dot1p classification policy with a SAP ingress policy.
The following example shows the command usage to associate a dot1p classification policy with a SAP ingress policy.
Use the following syntax to configure the classification policy to classify traffic to a forwarding class.
The following example shows the command usage to configure the classification policy to classify traffic to a forwarding class.
9.3.5.2. Associating a classification policy with an Ethernet port
Users associate an IP DSCP classification policy with an Ethernet port using the configure>port>ethernet>access>ingress>dscp-classification command.
Users associate a dot1p classification policy with an Ethernet port using the configure>port>ethernet>access>ingress>dot1p-classification command.
Use the following syntax to associate an IP DSCP classification policy with an Ethernet port.
The following example shows the command usage to associate an IP DSCP classification policy with an Ethernet port.
Use the following syntax to associate a dot1p classification policy with an Ethernet port.
9.3.6. Assigning and enabling DSCP and dot1p classification policies to a SAP
The 7210 SAS-Mxp supports table-based classification to assign an initial FC and profile on SAP ingress for epipe and VPLS SAPs, VPRN and IES interface SAPs, and RVPLS SAPs. The use of table-based classification (IP DSCP or dot1p) and CAM-based classification are mutually exclusive. That is, when table-based classification is used, any CAM-based classification configured from the SAP ingress QoS policy is ignored.
9.3.6.1. Assigning and enabling policies to Epipe and VPLS SAPs
Within epipe and VPLS services, SAPs can be configured with an IP DSCP or dot1p classification policy per SAP. This applies to SAPs configured on an access port and on a hybrid port. Using the enable-table-classification command means the SAP uses table-based policies along with the meters defined in the SAP ingress policy. CAM-based resources from the SAP ingress policy are ignored.
The enable-table-classification command enables the use of IP DSCP or dot1p tables per SAP ingress to assign an FC and profile. Using table-based classification means ignoring CAM classification in the service ingress policy, using only meters from service ingress policy, and using the IP DSCP classification policy or dot1p classification policy that is configured in the SAP ingress policy. The default FC is assigned per SAP.
The num-qos-classifiers command allocates meters from the IFP, with resources taken from the ingress-internal-tcam resource pool towards qos-sap-ingress.
The dscp-classification command configures which classification policy is used to match IP packets and to map an IP DSCP value to an FC and profile.
The dot1p-classification command configures which classification policy is used to match IP packets and to map a dot1p value to an FC and profile.
The table-classification-criteria command provides an option for all traffic to use either dot1p classification, or DSCP classification, or both IP DSCP and Dot1p classification, or assign default-fc (none option) to all traffic. The default-fc command configures the FC assigned to all untagged packets. All untagged packets are mapped to the profile “out”.
MAC, IPv4, and IPv6 criteria do not need to be defined because DSCP or dot1p classification is being used.
The CLI syntax below shows how to enable table classification for an Epipe and a VPLS service.
The following is a sample SAP ingress QoS policy 1000 configured with DSCP classification policy 101 and default-fc “be”. It is followed by a sample epipe ingress SAP configured to use SAP ingress policy 1000 with table-based classification enabled.
9.3.6.2. Assigning and enabling policies to IES and VPRN interface SAPs
Within IES and VPRN services, SAPs can be configured with an IP DSCP classification policy per SAP. This applies to SAPs configured on an access port and on a hybrid port. Using the enable-table-classification command means the SAP uses table-based policies along with the meters defined in the SAP ingress policy.
The enable-table-classification, num-qos-classifiers, dscp-classification, and default-fc commands for IES and VPRN interface SAPs operate similarly to Epipe and VPLS SAPs (see Assigning and enabling policies to Epipe and VPLS SAPs).
MAC, IPv4, and IPv6 criteria do not need to be configured because DSCP classification is being used.
Use the following syntax to enable table-based classification for an IES and a VPRN service.
The following is a sample SAP ingress QoS policy 1000 configured with DSCP classification policy 101 and default-fc “be”. It is followed by a sample ingress SAP on an IES service interface configured to enable table-based classification using SAP ingress policy 1000.
9.3.6.3. Assigning policies to RVPLS SAPs
For RVPLS SAPs configured on an access port, the 7210 SAS-Mxp supports RVPLS service with per-port IP DSCP classification policies or dot1p classification policies for bridged traffic received on SAPs configured in the RVPLS service. For routed traffic, per-IP interface IP DSCP or dot1p classification policies (that is, the QoS override policy) are used.
For RVPLS SAPs configured on a hybrid port, the network QoS policy of type “port” associated with network port ingress is used for RVPLS SAP bridged traffic classification and profile. Only the traffic classification will be used from the network policy. Meters are still used from the SAP ingress policy attached to the RVPLS SAP. For routed traffic received on a hybrid port, the IP DSCP or dot1p policy (that is, the QoS override policy) associated with the RVPLS IP interface is used for traffic classification and profile.
Only meters configured in the SAP ingress policy associated with RVPLS SAPs are used when table-based classification is enabled under the SAP associated with an RVPLS service.
| Note: All SAPs in an RVPLS service can use either table-based classification or CAM-based entries. There cannot be a mix of SAPs, where some SAPs use table-based classification and others use CAM entries. |
9.3.6.3.1. Create a SAP ingress policy and assign the policy to an RVPLS SAP
The following examples create and assign a SAP ingress QoS policy to an RVPLS SAP. Table-based classification is enabled in the override policy associated with the IES interface that is associated with the RVPLS service. In this case, only meters from the SAP ingress QoS policy are used. Ingress CAM entries are ignored (not used).
9.3.6.3.2. Table-based classification per IP Interface for routed packets
For routed packets, although the IP DSCP classification is based on the DSCP policy that is attached to the IP interface, the enable-table-classification command must also be set on RVPLS SAPs for table-based classification to work correctly. If enable-table-classification is not set on an RVPLS SAP, only the profile will be taken from the routed-override-qos policy for that SAP. In this case, traffic classification (in accordance with TCAM-based classification) and meters will be taken from the SAP ingress policy attached to the RVPLS SAP.
The following syntax enables table-based classification and specifies the QoS override classification policy in the IES or VPRN interface RVPLS configuration. The policy-id specified in the routed-override-qos-policy command identifies the DSCP policy configured using the configure>qos>dscp-classification command.
9.3.6.3.3. Table-based classification per port for bridged packets
For bridged packets, although the DSCP classification is based on the DSCP policy attached to the port, the enable-table-classification command must also be set in the IES or VPRN interface context as well as the respective RVPLS SAP context for table-based classification to work correctly (as shown in the Example). If enable-table-classification is not set on the respective RVPLS SAP then only profile will be taken from the port policy for that SAP. In this case, classification (in accordance with TCAM-based classification) and meters will be taken from the SAP ingress policy.
The following syntax enables table-based classification on an Ethernet port and specifies the DSCP classification policy in the port>ethernet>access>ingress command.
9.4. Service meter for SAP ingress (7210 SAS-Mxp)
Service meters for SAP ingress provide an option to use meter resources from the ingress service-meter pool, which provides a larger number of meters/policers for use by access SAPs. This option is available only with table-based classification; it is not available when CAM-based classification is used.
| Note: Table-based classification uses meters from either the TCAM pool or the service meter pool, based on the SAP ingress policy type. If the SAP ingress policy is configured to use the use-svc-meter-pool parameter, the policy uses the service meter pool, otherwise the policy uses the TCAM meter pool. |
Access to larger numbers of meters/policers from the service-meter pool is useful when there is a need to enforce bandwidth limits for all the FCs and traffic-types (that is, unicast, broadcast, multicast, and unknown-unicast) across a large number of access SAPs. The following functionality is available with the service-meter option:
- All access SAPs that use the service-meter pool share a single FC-to-meter-map policy that defines the meter to use for the configured FCs (there is a single configurable policy other than the default FC-to-meter map).
- The SAP ingress policy must be configured to use table-based classification with both IP DSCP and dot1p classification policies configured.
- SAP ingress policies can be configured with different IP DSCP and dot1p classification policies, which provide flexibility when configuring mappings for IP DSCP and dot1p values to FC for different SAPs.
- Even though all access SAPs share the same fc-to-meter-map policy, the meter instances and counter instances are different for each SAP that is configured. This means that each SAP gets an individual instance of meter to enforce rate-limits for traffic that is received on the SAP.
The following usage restrictions apply to meters across FCs and traffic types:
- If a meter is mapped to one or more traffic types (unicast, broadcast, multicast, and unknown-unicast) in a specific FC, the meter must be used with the same mapping in other FCs.
- Resources are always allocated in chunks of 1, 2, 4, 8, 16, or 32 meters. Counters are allocated in chunks of power of 2.
| Note: Service meter is supported for access SAPs that are configured in Epipe, VPLS, IES, and VPRN services. Service meter is not supported for access SAPs that are configured in an RVPLS service. |
9.4.1. Default service meter policy
The following is a sample default service meter policy output.
9.4.2. Resource usage for service meters
The number of meters used for a SAP ingress policy that is configured to use the service-meter pool is equal to the number of unique meters mapped to different FCs in the fc-meter-map policy that is associated with the SAP ingress policy.
The fc-meter-map defines the association between the meters and the FC-to-meter. Meter resources are allocated only to meters that have an associated FC. For example, if 10 meters are created in the fc-meter-map, and only 5 meters are associated with an FC, the system allocates only 5 meters per SAP (and not 10 meters) and rounds off the number of meters to the nearest power of 2, which results in eight (8) meters to be allocated. Meters that are not associated with an FC are ignored for resource allocation. The number of counters allocated is equal to twice the number of meters per SAP.
| Note:
|
9.4.2.1. Examples for service meters with computation of resource usage
The examples in this section use the following terminology:
- fc-name (Ucast) - unicast meter that is used for FC and specified by the fc-nameFor example, to configure a unicast meter that uses the FC value “be,” the notation be(Ucast) is used.
- B - broadcast meter
- U - unknown-unicast meter (also referred to as unknown meter)
- M - multipoint meter
This section provides examples of usage for service meters.
Example 1: Single FC configured to use the default unicast meter for all traffic (both unicast and BUM traffic)
Example 2: Single FC configured to use the default unicast meter for all unicast traffic and default multipoint meter (meter 11) for all BUM traffic
Example 3: Two FCs configured to use the default unicast meter for all unicast traffic and default multipoint meter (meter11) for all BUM traffic
Example 4: Two FCs configured to use the default unicast meter for all unicast traffic, with FC “be” configured to use default multipoint meter (meter 11) for all BUM traffic, with FC “l2” configured to use an explicit broadcast meter only broadcast traffic and the default multipoint meter 11 for both multicast and unknown-unicast traffic
Example 5: Two FCs configured, with one FC (that is, FC “be”) to use the explicitly configured unicast meter for unicast traffic and explicit meter for BUM traffic, and the other FC (that is, FC “l2”) configured to use default unicast meter for unicast traffic and explicit meter for BUM traffic
Example 6: One FC (that is, FC “be”) configured to use the explicitly configured unicast meter for all unicast traffic and BUM traffic
9.5. Calculating resources required for classification
This section provides examples for calculating the resources required for SAP-ingress policy classification when using CAM-based classification and table-based classification.
9.5.1. Examples: calculating resources required for CAM-based classification
This section provides examples for calculating the amount of resources needed for a service ingress policy when CAM-based classification is used. For calculations when IP DSCP table-based classification is used, see Examples: calculating resources required for IP DSCP table-based classification with CAM-based policing (7210 SAS-Mxp, 7210 SAS-R6, and 7210 SAS-R12).
The resource calculation shown for VLL is also applicable for VPRN services.
The examples in this section use the two equations below to calculate the value for num-qos-classifiers used in the sap-ingress QoS policy. See Computation of resources used per SAP ingress policy for CAM-based classification for more information about these equations.
- total number of classification entries (TC)TC = E(i) * C(i)i = nc, h1, ef, h2, l1, af, l2, bewhere:
- E(i) is the number of match-criteria entries that classify packets to FCi
- C(i) is the number of classification entries that are required by FCi to identify different traffic types
- total number of QoS resources required (TQ)TQ = max [(TC), (2 * TP))where:
- TP is the number of meters/policers used
In addition, the examples show how to determine the number of classification entries for each forwarding class. For example, FCh2 (shown below) is the sum of four traffic types: (unicast (U), broadcast (B), multicast (M), and unknown-unicast (U-u)). See Calculating the number of classification entries per FC for more information.
If BUM entries are not explicit and multipoint traffic is expected, meter "11" is used and the "M" traffic type is given a "1".
9.5.1.1. Example 1
In the preceding example, assuming the policy is attached to a SAP in a VPLS service, compute the number of classification entries per FC as follows.
Since FCh2 uses unicast meter, an entry is needed to identify unicast traffic type explicitly. Another entry is needed to classify broadcast, multicast and unknown-unicast traffic type to the same FC and use the default meter “11”.
Using the equation, calculate the total number of classification entries (TC) used by this policy, as follows:
TC = (0 * 0)nc + (0 * 0)h1 + (0 * 0)ef + (1 * 2)h2 + (1 * 2)l1 + (1 * 2)af + (0 * 0)l2 + (1 * 2)be = 8 (since three explicit match criteria entries are used to map traffic to each of FC H2, FC L1, and FC AF along with a default classification entry for FC BE).
The total number of meters used = 3 (since FCs use meter “1”, meter “3” and meter “11”).
In this example, num-qos-classifiers 8 is used (maximum of (8, (2 * 3))).
If the same policy were to be used for a SAP in an Epipe service, then since all traffic is classified to a unicast traffic type and since only unicast meters are used, the following computation is made:
Using the above equation, total classification entries used = 4 and meters used = 2.
As can be seen here, using the same policy for Epipe SAP can lead to inefficient use of resources. Hence, it is recommended to create a different policy with the required number of resources (that is, with num-qos-classifiers 4)
9.5.1.2. Example 1a (default multipoint meter 11 is not used):
In the example above, assuming the policy is attached to a SAP in a VPLS service, compute the number of classification entries per FC as follows:
Since FCh2 uses unicast meter for all traffic types, we need an entry to classify all traffic types to FCh2 explicitly.
Using the equation, calculate the total classification entries used by this policy, as follows:
TC = (0 * 0)nc + (0 * 0)h1 + (0 * 0)ef + (1 * 1)h2 + (1 * 1)l1 + (1 * 1)af + (0 * 0)l2 + (1 *1)be = 4
(since three explicit match criteria entries are used to map traffic to each of FC H2, FC L1, and FC AF along with a default classification entry for FC BE).
The total number of meters used = 2 (since FCs use meter “1” and meter “3”).
In this example, num-qos-classifiers 4 is used (maximum of (4, (2 * 2))). Use of unicast meter for all traffic-types allows for use QoS resources efficiently.
If the same policy were to be used for a SAP in an Epipe service, then since all traffic is classified to a unicast traffic type and since only unicast meters are used, the following:
Using the above equation for TC calculation, total classification entries used = 4 and meters used = 2.
As can be seen here, using the same policy for Epipe SAP can lead to inefficient use of resources. Hence, it is recommended to create a different policy with the required number of resources (that is, with num-qos-classifiers 4).
9.5.1.3. Example 2
In the preceding example, assuming the policy is attached to a SAP in a VPLS service, classification entries used per FC as:
Since FCh2 uses unicast meter and broadcast meter, two entries are needed to identify these traffic types explicitly. Another entry is needed to classify multicast and unknown-unicast traffic type to FCh2 and use the default meter “11”.
Using the above equation for TC calculation, to get the total classification entries used = 12 (since three explicit match criteria entries map to each of FC H2, L1, and AF along with a default classification rule for BE).
The number of meters used = 3 (since FCs use only meter “2”, meter “3” and meter “11”).
Hence, in this example num-qos-classifiers 16 is used (i.e., maximum of (12, (2*3))).
If the same policy were to be used for a SAP in an Epipe service, then since all traffic is classified to a unicast traffic type and since only unicast meters are used, the following:
Using the above equation, to get total classification entries used = 4 and Meters used = 1.
As can be seen here, using the same policy for Epipe SAP can lead to inefficient use of resources. Hence, it is recommended to create a different policy with the required number of resources (i.e. with num-qos-classifiers 4)
9.5.1.4. Example 2a (default multipoint meter "11" is not used)
In the example above, assuming the policy is attached to a SAP in a VPLS service, classification entries used per FC as:
Since FCh2 uses unicast meter for unicast, multicast, and unknown-unicast traffic, and broadcast meter for broadcast traffic, two entries are needed.
Using the above equation, to get the total classification entries used = 8 (since three explicit match criteria entries map to each of FC H2, L1, and AF along with a default classification rule for BE).
The number of meters used = 2 (since FCs use only meter “2” and meter “3”).
Hence, in this example num-qos-classifiers 8 is used (that is, maximum of (8, (2*2))).
If the same policy were to be used for a SAP in an Epipe service, then since all traffic is classified to a unicast traffic type and since only unicast meters are used, the following:
Using the above equation, to get total classification entries used = 4 and Meters used = 1.
As can be seen here, using the same policy for Epipe SAP can lead to inefficient use of resources. Hence, it is recommended to create a different policy with the required number of resources (that is, with num-qos-classifiers 4)
9.5.1.5. Example 3
In the example above, assuming the policy is attached to a SAP in a VPLS service, the classification entries used per FC are:
Since FCh1 uses unicast, broadcast, multicast and unknown-unicast meter, four entries are needed to identify these traffic types explicitly.
Since FCh2 uses unicast meter and broadcast meter, two entries are needed to identify these traffic types explicitly. Another entry is needed to classify multicast and unknown-unicast traffic type to the same FC and use the default meter “11”.
Since FCl1 uses only unicast meter, an entry is needed to identify this traffic type explicitly. Another entry is needed to classify broadcast, multicast and unknown-unicast traffic type to the same FC and use the default meter “11”.
Since FCaf uses unicast, broadcast and multicast meter, three entries are needed to identify these traffic types explicitly. Unknown-unicast traffic type is classified using the same entry as multicast traffic type and uses the same meter.
Using the above equation, the total classification entries used = 15 and meters used = 6.
If the same policy were to be used for a SAP in an Epipe service, then since all traffic is classified to a unicast traffic type and since only unicast meters are used, the following results:
Using the above equation, the total classification entries used = 5 and meters used = 3 (since all FCs used only meter “1”, meter “3” and meter “5”).
9.5.1.6. Example 3a (default multipoint meter "11" is not used):
In the example above, assuming the policy is attached to a SAP in a VPLS service, the classification entries used per FC are:
Since FCh1 uses unicast, broadcast, multicast and unknown-unicast meter, four entries are needed to identify these traffic types explicitly.
Since FCh2 uses unicast meter and broadcast meter, two entries are needed to identify these traffic types explicitly, multicast and unknown-unicast traffic use the same resource as the unicast traffic.
Since FCl1 uses unicast meter and broadcast meter, two entries are needed to identify these traffic types explicitly. multicast and unknown-unicast traffic use the same resource as the unicast traffic.
Since FCaf uses unicast, broadcast and multicast meter, three entries are needed to identify these traffic types explicitly. Unknown-unicast traffic type is classified using the same entry as multicast traffic type and uses the same meter.
Since no explicit meters are configured for FC "be", it uses meter "1" for all traffic types and needs one entry is needed to identify these traffic types.
Using the above equation, the total classification entries used = 12 and meters used = 5. The num-qos-classifiers can be set to 12 (the minimum value).
If the same policy were to be used for a SAP in an Epipe service, then since all traffic is classified to a unicast traffic type and since only unicast meters are used, the following results:
Using the above equation, the total classification entries used = 5 and meters used = 3 (since all FCs used only meter “1”, meter “3” and meter “5”). For Epipe service a policy with num-qos-classifiers set to 6 can be used.
9.5.1.7. Example 4
In the example above, assuming the policy is attached to a SAP in a VPLS service, compute the classification entries per FC as:
Since FCnc uses unicast, broadcast and multicast meter, three entries are needed to identify these traffic types explicitly. Unknown-unicast traffic type is classified using the same entry as multicast traffic type and uses the same meter.
Since FCh1 uses unicast, broadcast, multicast and unknown-unicast meter, four entries are needed to identify these traffic types explicitly.
Since FCef uses unicast, broadcast and multicast meter, three entries are needed to identify these traffic types explicitly. Unknown-unicast traffic type is classified using the same entry as multicast traffic type and uses the same meter.
Since FCh2 uses unicast meter and broadcast meter, two entries are needed to identify these traffic types explicitly. Another entry is needed to classify multicast and unknown-unicast traffic type to the same FC and use the default meter “11”.
Since FCaf uses unicast, broadcast and multicast meter, three entries are needed to identify these traffic types explicitly. Unknown-unicast traffic type is classified using the same entry as multicast traffic type and uses the same meter.
Using the above equation, the total classification entries used = 21 and meters used = 8.
If the same policy were to be used for a SAP in an Epipe service, then since all traffic is classified to a unicast traffic type and since only unicast meters are used, the following:
Using the above equation, the total classification entries used = 7 and meters used = 4.
As can be seen here, using the same policy for Epipe SAP can lead to inefficient use of resources. Hence, it is recommended to create a different policy with the required number of resources (that is, with num-qos-classifiers 8)
9.5.1.8. Example 4a (default multipoint meter "11" is not used):
In the example above, assuming the policy is attached to a SAP in a VPLS service, compute the classification entries per FC as:
Since FCnc uses unicast, broadcast and multicast meter, three entries are needed to identify these traffic types explicitly. Unknown-unicast traffic type is classified using the same entry as multicast traffic type and uses the same meter.
Since FCh1 uses unicast, broadcast, multicast and unknown-unicast meter, four entries are needed to identify these traffic types explicitly.
Since FCef uses unicast, broadcast and multicast meter, three entries are needed to identify these traffic types explicitly. Unknown-unicast traffic type is classified using the same entry as multicast traffic type and uses the same meter.
Since FCh2 uses unicast meter and broadcast meter, two entries are needed to identify these traffic types explicitly. multicast and unknown-unicast traffic of the same FC use the unicast resources (both meter and classification entry).
Since FCaf uses unicast, broadcast and multicast meter, three entries are needed to identify these traffic types explicitly. Unknown-unicast traffic type is classified using the same entry as multicast traffic type and uses the same meter.
Since FCbe uses a single meter for all traffic-types only a single meter and single entry is needed.
Using the above equation, the total classification entries used = 20 and meters used = 7, num-qos-classifiers to use is 20 (the minimum value).
If the same policy were to be used for a SAP in an Epipe service, then since all traffic is classified to a unicast traffic type and since only unicast meters are used, the following:
Using the above equation, the total classification entries used = 7 and meters used = 4.
As can be seen here, using the same policy for Epipe SAP can lead to inefficient use of resources. Hence, it is recommended to create a different policy with the required number of resources (that is, with num-qos-classifiers 8).
9.5.1.9. Example 5
In the example above, assuming the policy is attached to a SAP in a VPLS service, get the classification entries used per FC:
Since FCnc uses unicast, broadcast and multicast meter, three entries are needed to identify these traffic types explicitly. Unknown-unicast traffic type is classified using the same entry as multicast traffic type and uses the same meter.
Since FCh1 uses unicast, broadcast, multicast and unknown-unicast meter, four entries are needed to identify these traffic types explicitly.
Since no meters are explicitly configured, FCef uses the appropriate default meters all the traffic types (i.e. unicast traffic uses unicast meter “1” and broadcast, multicast, and unknown-unicast traffic uses multipoint meter “11”.
Since FCh2 uses unicast meter and broadcast meter, two entries are needed to identify these traffic types explicitly. Another entry is needed to classify multicast and unknown-unicast traffic type to the same FC and use the default meter “11”.
Since FCaf uses unicast, broadcast and multicast meter, three entries are needed to identify these traffic types explicitly. Unknown-unicast traffic type is classified using the same entry as multicast traffic type and uses the same meter.
Using the above equation, the total classification entries used = 20 and meters used = 8.
If the same policy were to be used for a SAP in an Epipe service, then since all traffic is classified to a unicast traffic type and since only unicast meters are used, the following:
Using the above equation, to get the total classification entries used = 7 and meters used = 4.
9.5.1.10. Example 6
In the example above, assuming the policy is attached to a SAP in a VPLS service, the following number of classification entries per FC:
Since FCl1 uses unicast meter and multicast meter, an entry is needed to identify these traffic types explicitly. Broadcast and unknown-unicast traffic is also classified using the same entry as multicast and use the same meter.
Since FCaf uses unicast meter, an entry is needed to identify these traffic types explicitly. Another entry is needed to classify broadcast, multicast and unknown-unicast traffic type to the same FC and use the default meter “11”.
Using the above equation, the total classification entries used = 8 and meters used = 4.
If the same policy were to be used for a SAP in an Epipe service, then since all traffic is classified to a unicast traffic type and since only unicast meters are used, the following:
Using the above equation, the total classification entries used = 4 and meters used = 2.
9.5.1.11. Example 7
In the example above, assuming the policy is attached to a SAP in a VPLS service, the following number of classification entries per FC are:
Since FCaf uses unicast meter, an entry is needed to identify these traffic types explicitly. Another entry is needed entry to classify broadcast, multicast and unknown-unicast traffic type to the same FC and use the default meter “11”.
Since FCbe uses unicast, broadcast and multicast meter, three entries are needed to identify these traffic types explicitly. Unknown-unicast traffic type is classified using the same entry as multicast traffic type and uses the same meter.
Using the above equation, the total classification entries used = 5 and meters used = 4.
If the same policy were to be used for a SAP in an Epipe service, then since all traffic is classified to a unicast traffic type and since only unicast meters are used, the following:
Using the above equation, the total classification entries used = 2 and meters used = 2.
9.5.1.12. Example 8
In the example above, assuming the policy is attached to a SAP in a VPLS service, the following number of classification entries per FC:
Since FCaf uses unicast meter, an entry is needed to identify these traffic types explicitly. Another entry is needed to classify broadcast, multicast and unknown-unicast traffic type to the same FC and use the default meter “11”.
Since FCbe uses unicast, broadcast and multicast meter, three entries are needed to identify these traffic types explicitly. Unknown-unicast traffic type is classified using the same entry as multicast traffic type and uses the same meter.
Using the equation, calculate the total classification entries used by this policy, as follows
TC = (0 * 0)nc + (0 * 0)h1 + (0 * 0)ef + (0 * 0)h2 + (0 * 0)l1 + (3 * 2)af + (0 * 0)l2 + (1 * 3)be = 9
The number of meters used in this policy = 4.
Hence, in this example num-qos-classifiers 16 is used (i.e. maximum of (9, (2 * 4))).
If the same policy were to be used for a SAP in an Epipe service, then since all traffic is classified to a unicast traffic type and since only unicast meters are used, the following:
Using the equation, calculate the total classification entries used by this policy, as follows:
TC = (0 * 0)nc + (0 * 0)h1 + (0 * 0)ef + (0 * 0)h2 + (0 * 0)l1 + (3 * 1)af + (0 * 0)l2 + (1 * 1)be = 4
The number of meters used in this policy = 2.
9.5.1.13. Example 9
In the example above, assuming the policy is attached to a SAP in a VPLS service, the following number of classification entries per FC:
Since FCaf uses unicast meter, an entry is needed to identify these traffic types explicitly. Another entry is needed to classify broadcast, multicast and unknown-unicast traffic type to the same FC and use the default meter “11”.
Since FCbe uses unicast, broadcast and multicast meter, three entries are needed to identify these traffic types explicitly. Unknown-unicast traffic type is classified using the same entry as multicast traffic type and uses the same meter.
Using the equation, calculate the total classification entries used by this policy, as follows:
TC = (0 * 0)nc + (0 * 0)h1 + (0 * 0)ef + (0 * 0)h2 + (0 * 0)l1 + (50 * 3)af + (0 * 0)l2 + (1 * 3)be = 153
The number of meters used in this policy = 4.
Hence, in this example num-qos-classifiers 256 is used (maximum of (153, (2 * 4)) = 153, rounded off to the next multiple of 2 will be 154).
If the same policy were to be used for a SAP in an Epipe service, then since all traffic is classified to a unicast traffic type and since only unicast meters are used, e the following:
Using the equation, calculate the total classification entries used by this policy, as follows:
TC = (0 * 0)nc + (0 * 0)h1 + (0 * 0)ef + (0 * 0)h2 + (0 * 0)l1 + (50 * 1)af + (0 * 0)l2 + (1 * 1)be = 51
The number of meters used in this policy = 2.
Hence for Epipe SAP it is recommended to define another sap-ingress policy with num-qos-classifiers 64 is used (i.e. maximum of (51, (2 * 2)) = 51, rounded off to the next multiple of 2 will be 52).
9.5.1.14. Example 9a (default multipoint meter "11" is not used):
In the example above, assuming the policy is attached to a SAP in a VPLS service, the following number of classification entries per FC:
Since FCaf uses unicast, broadcast and multicast meter, three entries are required to identify these traffic types explicitly. Unknown-unicast traffic type is classified using the same entry as multicast traffic type and uses the same meter. Additionally note that meter "11" is not defined to be multipoint meter, but is used as a normal unicast meter.
Since FCbe uses unicast, broadcast and multicast meter, three entries are needed to identify these traffic types explicitly. Unknown-unicast traffic type is classified using the same entry as multicast traffic type and uses the same meter. Additionally note that meter "11" is not defined to be multipoint meter, but is used as a normal unicast meter.
Using the equation, calculate the total classification entries used by this policy, as follows:
TC = (0 * 0)nc + (0 * 0)h1 + (0 * 0)ef + (0 * 0)h2 + (0 * 0)l1 + (50 * 3)af + (0 * 0)l2 + (1 * 3)be = 153
The number of meters used in this policy = 4. Hence, in this example num-qos-classifiers 154 is used (maximum of (153, (2 * 4)) = 153, rounded off to the next multiple of 2 will be 154).
Hence, in this example num-qos-classifiers 154 is used (maximum of (153, (2 * 4)) = 153, rounded off to the next multiple of 2 will be 154).
If the same policy were to be used for a SAP in an Epipe service, then since all traffic is classified to a unicast traffic type and since only unicast meters are used, the following:
Using the equation, calculate the total classification entries used by this policy, as follows:
TC = (0 * 0)nc + (0 * 0)h1 + (0 * 0)ef + (0 * 0)h2 + (0 * 0)l1 + (50 * 1)af + (0 * 0)l2 + (1 * 1)be = 51
The number of meters used in this policy = 2.
Hence for Epipe SAP it is recommended to define another sap-ingress policy with num-qos-classifiers 52 is used (that is, maximum of (51, (2 * 2)) = 51, rounded off to the multiple of 2 will be 52).
9.5.1.15. Example 10
In the example above, assuming the policy is attached to a SAP in a VPLS service, compute the number of classification entries per FC as follows:
Using the equation, calculate the total classification entries used by this policy, as follows:
TC = (0 * 0)nc + (0 * 0)h1 + (0 * 0)ef + (0 * 0)h2 + (0 * 0)l1 + (1 * 2)af + (1 * 2)l2 + (0 * 0)be = 4
The number of meters used = 2 (since both FCs use meter “1” and meter “11”).
Hence, in this example num-qos-classifiers 4 is used (that is, maximum of (4, (2 * 2))).
If the same policy were to be used for a SAP in an Epipe service, then since all traffic is classified to a unicast traffic type and since only unicast meters are used, the following:
Using the above equation, calculate the total classification entries used = 2 and meters used = 1.
As can be seen here, for Epipe SAP with the same amount of resources allocated one can have more FCs if need be.
9.5.1.16. Example 11
In the example above, assuming the policy is attached to a SAP in a VPLS service, compute the number of classification entries per FC as follows:
Using the equation, calculate the total classification entries used by this policy, as follows:
TC = (0 * 0)nc + (0 * 0)h1 + (0 * 0)ef + (0 * 0)h2 + (0 * 0)l1 + (0 * 0)af + (1 * 2)l2 + (0* 0)be = 2
The number of meters used = 2 (since default FC uses meter “1” and meter “11”).
Hence, in this example num-qos-classifiers 4 is used (i.e. maximum of (2, (2 * 2))).
If the same policy were to be used for a SAP in an Epipe service, then since all traffic is classified to a unicast traffic type and since only unicast meters are used, the following:
Using the above equation, total classification entries used = 1 and meters used = 1.
As can be seen here, for Epipe SAP with the same amount of resources allocated one can have more FCs if need be.
9.5.2. Examples: calculating resources required for IP DSCP table-based classification with CAM-based policing (7210 SAS-Mxp, 7210 SAS-R6, and 7210 SAS-R12)
This section provides examples for calculating the amount of resources needed for a service ingress policy when using IP DSCP table-based classification with CAM-based policing. For calculations when CAM-based classification is used, see Examples: calculating resources required for CAM-based classification.
The examples in this section use the two equations below to calculate the value for num-qos-classifiers used in the sap-ingress QoS policy. See Computation of resources used per SAP ingress policy for CAM-based classification for details on these equations.
- total number of classification entries (TC)TC = E(i) * C(i)i = nc, h1, ef, h2, l1, af, l2, be, default-fcwhere:
- E(i) is equal to 1 if FC(i) is in use by the dscp-classification policy. Otherwise, E(i) is equal to 0.
- C(i) is the number of classification entries that are required by FCi to identify different traffic types. For a given FC, for each traffic type (unicast, broadcast, multicast, and unknown-unicast) configured to use a meter, a classification entry is needed.
Note: The default FC requires one or more additional resources even if all the eight FCs are configured in the dot1p or IP DSCP classification policy, as shown in the example in section 9.5.2.1.
- total number of QoS resources required (TQ)TQ = max [(TC), (2 * TP))where:
- TP is the number of meters/policers used
In addition, the examples show how to determine the number of classification entries for each forwarding class. For example, FCh2 (shown below) is the sum of four traffic types: (unicast (U), broadcast (B), multicast (M), and unknown-unicast (U-u)). See Calculating the number of classification entries per FC for more information.
If BUM entries are not explicit and multipoint traffic is expected, meter "11" is used and the "M" traffic type is given a "1".
Consider the following items when calculating the resources required when using IP DSCP table-based classification:
- The meters used per FC for different traffic types is in accordance with the rules given in Service ingress meter selection rules and the number of classification entries per FC is provided in Determining the number of classification entries. In addition:
- If an FC uses one meter for all four traffic types, then the maximum of four classification entries are needed.
- As a minimum, an FC uses a single meter.
- Users are provided an option to use between one and four meters per FC, one meter each for four different traffic types, in a VPLS service.
- Users are provided an option to use either one and two meters per FC, one meter each for two traffic types, in IES and VPRN service when using multicast.
- Users can use only a single meter per FC, for unicast traffic, in Epipe, IES, and VPRN service, where IES and VPRN are not configured to use multicast.
- For an Epipe, VPLS, IES, or VPRN SAP, the FCs in use can be determined by counting the FCs configured in the DSCP classification policy and the default FC configured in the SAP ingress policy.
- For routed VPLS (RVPLS), to determine the FCs in use, use the following:
- For SAPs on access port, the FCs in use is the sum total of all FCs configured across the following four items:
- the DSCP classification policy associated as an override policy under the IP interface context
- the port DSCP classification policy associated with the access port
- the untagged-fc value configured under the access port
- the default-fc configured in the SAP ingress policy
- For SAPs on hybrid ports, the FCs in use is the sum total of all FCs configured across the following three items:
- the DSCP classification policy associated as an override policy under the IP interface context
- the network port policy associated with the hybrid port
- the default FC value configured under the SAP ingress policy
9.5.2.1. Example 1: Epipe, IES, and VPRN services using unicast traffic type
In the above example, all eight FCs are configured and eight meters are configured. For Epipe service only unicast traffic-type is identified. This requires one classification entry per FC configured and an additional one for the default-fc assignment, for a total of 9 classification entries.
TC = 1*1 (FC-nc) + 1*1 (FC-h1) + 1*1 (FC-ef) + 1*1 (FC-h2) + 1*1 (FC-l1) + 1*1 (FC-af) + 1*1 (FC-l2) + 1*1 (FC-be) + 1*1 (default-fc) = 9
TP = Total meters used is 8.
Hence, in this example num-qos-classifiers is set to (maximum (9, (8*2)) = 16.
If the same policy was going to be attached to an IES or VPRN SAP without multicast enabled (VPRN example shown below, IES is similar), the num-qos-classifiers required would be the same as that for am Epipe service (that is, 16). The calculations are the same as that for Epipe service.
9.5.2.2. Example 2: VPLS using unicast and BUM meter with IES or VPRN using multicast
In the above example all eight FCs are configured and eight meters are configured. In addition, multipoint meter "11" is configured for use. For the VPLS service, four traffic types are identified (unicast and BUM). Since multicast meter "11" is defined, BUM traffic type for all FCs will use meter "11". The number of classification entries required is:
TC = 1*2 (FC-nc) + 1*2 (FC-h1) + 1*2 (FC-ef) + 1*2 (FC-h2) + 1*2 (FC-l1) + 1*2 (FC-af) + 1*2 (FC-l2) + 1*2 (FC-be) + 1*2 (default-fc) = 18
TP = Total meters used is 8.
Hence, in this example num-qos-classifiers is set to (maximum (18, (8*2)) = 18.
If the same policy is attached to an IES or an VPRN service with multicast enabled, then the number of resources required would be 18 (that is, num-qos-classifiers needs to be set to 18). The calculations are the same those for VPLS service (shown above).
9.5.2.3. Example 3: VPLS service using unicast, broadcast, multicast, and unknown-unicast with additional FCs
In this example, the sap-ingress policy in Example 2 is changed to define additional meters (shown below) and applied to a SAP configured in a VPLS service.
For the VPLS service, four traffic types are identified (unicast and BUM). Since multicast meter "11" is defined, BUM traffic types for all FCs will use meter "11" (since the user has not configured an explicit multicast-meter for the FC). Hence, the number of classification entries required is determined as shown below:
TC = 1*2 (FC-nc) + 1*3 (FC-h1) + 1*3 (FC-ef) + 1*3 (FC-h2) + 1*4 (FC-l1) + 1*3 (FC-af) + 1*3 (FC-l2) + 1*2 (FC-be) + 1*2 (default-fc) = 25
TP = Total meters used is 11.
Hence, in this example, num-qos-classifiers is calculated to be (maximum (25, (11*2)) = 25, but is set to 26 after rounding off to the next highest even number.
If the same policy is attached to an IES or an VPRN service with multicast enabled, then the number of resources required would be 18 (that is, num-qos-classifiers needs to be set to 18). This is because for IES and VPRN only unicast and multicast traffic types are supported; broadcast and unknown-unicast traffic types are not supported and do not consume any resources. The calculation is as shown below:
TC = 1*2 (FC-nc) + 1*2 (FC-h1) + 1*2 (FC-ef) + 1*2 (FC-h2) + 1*2 (FC-l1) + 1*2 (FC-af) + 1*2 (FC-l2) + 1*2 (FC-be) + 1*2 (default-fc) = 18
TP = Total meters used is 8 (meter 20, meter 21, and meter 22 are not used since they are associated with broadcast and unknown-unicast traffic types, which are not supported for IES and VPRN).
Hence, in this example num-qos-classifiers is set to (maximum (18, (8*2)) = 18.
9.5.2.4. Example 4: routed VPLS on access port using unicast, broadcast, multicast, and unknown-unicast with additional FCs
IES service and routed VPLS service configuration is shown below.
The configuration for the DSCP classification policy associated with the access port that the RVPLS SAP is configured on, and which is used for classifying bridged packets is shown below:
For the bridged traffic in a VPLS service four traffic types are identified (unicast and BUM). Since multicast meter "11" is defined, BUM traffic type for all FCs will use meter "11" (Note that an explicit multicast-meter for the FC has not been configured). The number of classification entries required is as follows:
TC = 1*2 (FC-nc) + 1*3 (FC-h1) + 1*3 (FC-ef) + 1*3 (FC-h2) + 1*4 (FC-l1) + 1*3 (FC-af) + 1*3 (FC-l2) + 1*2 (FC-be) + 1*2 (default-fc) = 25
TP = Total meters used is 11.
Hence, in this example num-qos-classifiers is set to (maximum (25, (11*2)) = 25, which means 26 after rounding off to the next highest even number.
| Note: For routed traffic in the routed VPLS service only the unicast traffic type is supported currently. This does not change the amount of resources needed since bridged traffic requires higher amount of resources. To reduce the amount of resources, users can dedicate a single meter for BUM traffic from all FCs, as shown in the following example (note that meter "11" is used for all FCs automatically when meter "11" is defined in the policy). |
In the above example all eight FCs are configured and nine meters are configured, with multipoint meter "11" dedicated to all BUM traffic for all FCs in use. For the VPLS service four traffic-types are identified (unicast and BUM). Since multicast meter "11" is defined, BUM traffic type for all FCs will use meter "11". The number of classification entries required is as follows:
TC = 1*2 (FC-nc) + 1*2 (FC-h1) + 1*2 (FC-ef) + 1*2 (FC-h2) + 1*2 (FC-l1) + 1*2 (FC-af) + 1*2 (FC-l2) + 1*2 (FC-be) + 1*2 (default-fc) = 18
TP = Total meters used is 9.
Hence, in this example num-qos-classifiers is set to (maximum (18, (9*2)) = 18.
9.5.2.5. Example 5: routed VPLS service on a hybrid port using unicast, broadcast, multicast and unknown-unicast for some FCs
The IES service and routed VPLS service configuration is as follows:
The configuration for the network port policy for a hybrid port, followed by associating the policy with hybrid port 1/1/3 is shown below. This configuration is used for classifying RVPLS SAP bridged packets and also for classifying IP traffic received and processed in the context of the network port IP interface.
For the bridged traffic in a VPLS four traffic types are identified (unicast and BUM). Since multicast meter "11" is defined, BUM traffic type for all FCs will use meter "11" (Note that the user has not configured an explicit multicast-meter for the FC). The number of classification entries required is as follows.
TC = 1*2 (FC-nc) + 1*3 (FC-h1) + 1*3 (FC-ef) + 1*3 (FC-h2) + 1*4 (FC-l1) + 1*3 (FC-af) + 1*3 (FC-l2) + 1*2 (FC-be) + 1*2 (default-fc) = 25
TP = Total meters used is 11.
Hence, in this example num-qos-classifiers is calculated to be (maximum (25, (11*2)) = 25, but is set to 26 after rounding off to the next highest even number.
9.5.2.6. Example 6: routed VPLS on access port and hybrid port
Example 6 is similar to Example 5, except that FCs in use from the port policy and the override policy are considered. The following configuration shows the policy association with two SAPs configured in the routed VPLS, with one SAP on the access port and the other SAP on the hybrid port.
The following are sample IES service and routed VPLS service configuration outputs.
The configuration of the network port policy for a hybrid port, followed by associating the policy with hybrid port 1/1/3 is shown below. This configuration is used for classifying RVPLS SAP bridged packets and also for classifying IP traffic received and processed in the context of the network port IP interface.
The configuration of the DSCP classification policy associated with the access port, which is the port that the RVPLS SAP is configured on, is shown below. The policy is used for classifying bridged packets.
To determine the resources needed for RVPLS SAP 1/1/3:201 consider the FCs configured in the DSCP classification policy configured in access port 1/1/3 context—there are three FCs (be, af, ef) configured. In addition, consider the FCs configured in the DSCP classification policy configured in the context of IES IP interface—there are three FCs (be, af, ef) configured. The total number of resources for RVPLS SAP 1/1/3:201 is computed as follows using the meter configuration under SAP ingress policy 24.
TC = 0*2 (FC-nc) + 0*3 (FC-h1) + 1*3 (FC-ef) + 0*3 (FC-h2) + 0*4 (FC-l1) + 1*3 (FC-af) + 0*3 (FC-l2) + 1*2 (FC-be) + 1*2 (default-fc) = 10
TP = Total meters used is 4 (meter 11, meter 9, meter 22, and meter 14).
Hence, to use the available resources efficiently for the RVPLS SAP 1/1/3:201 we need a policy with num-qos-classifiers set to (maximum (10, (4*2)) = 10.
To determine the resources needed for RVPLS SAP 1/1/24:201 consider the FCs configured in the network port policy 123, which is associated with network port 1/1/24—all eight FCs are configured. In addition, consider the FCs configured in the DSCP classification policy configured in the context of IES IP interface—three FCs (be, af, ef) are configured. The total number of resources for RVPLS SAP 1/1/24:201 is computed as follows, using the meter configuration under the SAP ingress policy 24:
TC = 1*2 (FC-nc) + 1*3 (FC-h1) + 1*3 (FC-ef) + 1*3 (FC-h2) + 1*4 (FC-l1) + 1*3 (FC-af) + 1*3 (FC-l2) + 1*2 (FC-be) + 1*2 (default-fc) = 25
TP = Total meters used is 11.
Hence, for RVPLS SAP /1/124:201, we need a policy with num-qos-classifiers in the policy is calculated to be (maximum (25, (11*2)) = 25, but is set to 26 after rounding off to the next highest even number.
9.6. Basic configurations
A basic service ingress QoS policy must conform to the following:
- Have a unique service ingress QoS policy ID.
- Allocate number of classifier and meter resources needed for use
- Have a QoS policy scope of template or exclusive.
- Have at least one default unicast forwarding class meter/queue.
- (optionally) Use multipoint forwarding class meter/queues.
Configuring and applying QoS policies is optional. If no QoS policy is explicitly applied to a SAP, a default QoS policy is applied.
9.6.1. Create service ingress QoS policies
To create a service ingress Q0S policy, define the following:
- A policy ID value. The system will not dynamically assign a value.
- Include a description. The description provides a brief overview of policy features.
- Specify the num-qos-classifiers num-resources parameter. The default value is 2.
- Specify a default forwarding class for the policy. All packets received on an ingress SAP using this ingress QoS policy will be classified to the default forwarding class.
- Define forwarding class parameters:
- Modify the unicast-meter/queue default value to override the default unicast forwarding type meter mapping for fc fc-name.
- Modify the multicast-meter/queue default value to override the default multicast forwarding type meters/queue mapping for fc fc-name.
- Modify the unknown-meter/queue default value to override the default unknown unicast forwarding type meter mapping for fc fc-name.
- Modify the broadcast-meter default value to override the default broadcast forwarding type meter mapping for fc fc-name.
- On platforms where applicable, specify the appropriate classification criteria: IPv4/IPv6 or MAC criteria, or both IP and MAC criteria. You can define IPv4/IPv6, MAC-based, and MAC- and IP-based SAP ingress policies to select the appropriate ingress meter and corresponding forwarding class for matched traffic. See Service ingress IP match criteria and Service ingress MAC match criteria.
- A SAP ingress policy is created with a template scope. The scope can be modified to exclusive for a special one-time use policy. Otherwise, the template scope enables the policy to be applied to multiple SAPs.
The following is a sample service ingress policy configuration output.
9.6.1.1. Service ingress QoS meter
To create service ingress meter parameters, define the following:
- A new meter ID value — The system will not dynamically assign a value.
- Meter parameters — Ingress meters support the definition of either srTCM (Single Rate Tri-Color Meter) or trTCM (Two Rate Tri-Color Meter), CIR/PIR, CBS/MBS parameters.
The following is a sample ingress meter configuration output.
9.6.1.2. Service ingress IP match criteria
When specifying SAP ingress match criteria, only one match criteria type can be configured in the SAP ingress QoS policy.
The following are two sample ingress IP criteria configuration outputs.
9.6.1.3. Service ingress MAC match criteria
To configure service ingress QoS policy MAC criteria, define the following:
- a new entry ID valueEntries must be explicitly created. The system will not dynamically assign entries or a value.
- actionThe action to associate the forwarding class with a specific MAC criteria entry ID.
- a descriptionThe description provides a brief overview of policy features.
The following is a sample ingress MAC criteria configuration output.
9.6.2. Applying service ingress policies
This section describes applying SAP ingress policies to service SAPs.
9.6.2.1. Epipe
The following is a sample Epipe service configuration output with SAP ingress policy 100 applied to the SAP. The enable-table-classification keyword applies only to the 7210 SAS-Mxp, 7210 SAS-R6, and 7210 SAS-R12.
9.6.2.2. VPLS
The following is a sample VPLS service configuration output with SAP ingress policy 100. The enable-table-classification keyword applies only to the 7210 SAS-Mxp, 7210 SAS-R6, and 7210 SAS-R12.
9.6.2.3. VPRN
The following is a sample VPRN service configuration output. The enable-table-classification keyword applies only to the 7210 SAS-Mxp, 7210 SAS-R6, and 7210 SAS-R12.
9.6.2.4. IES
The following is a sample IES service configuration output. The enable-table-classification keyword applies only to the 7210 SAS-Mxp, 7210 SAS-R6, and 7210 SAS-R12.
9.7. Service management tasks
This section describes service management tasks.
9.7.1. Deleting QoS policies
Every service SAP is associated, by default, with the appropriate ingress policy (policy-id 1). You can replace the default policy with a customer-configured policy, but you cannot entirely remove the policy from the SAP configuration. When you remove a non-default service ingress policy, the association reverts to the default policy-id 1.
A QoS policy cannot be deleted until it is removed from all SAPs where they are applied.
9.7.1.1. Remove a QoS policy from service SAPs
The following Epipe service output examples show that the SAP service ingress reverted to policy-id “1” when the non-default policies were removed from the configuration.
9.7.2. Copying and overwriting QoS policies
You can copy an existing service ingress policy, rename it with a new policy ID value, or overwrite an existing policy ID. The overwrite option must be specified or an error occurs if the destination policy ID exists.
Use the following syntax to copy and overwrite QoS policies.
9.7.3. Remove a policy from the QoS configuration
Use the following syntax to remove a policy from the QoS configuration.
9.7.4. Editing QoS policies
You can change QoS existing policies and entries. The changes are applied immediately to all services where this policy is applied. To prevent configuration errors copy the policy to a work area, make the edits, and then write over the original policy.