QoS on access port LAGs (access ports and hybrid ports in access mode) is handled differently from QoS on network port LAGs (see QoS for LAG on Network). Based on the configured hashing, traffic on a SAP can be sent over multiple LAG ports or can use a single port of a LAG. There are two user-selectable adaptive QoS modes (distribute and link) that allow the user to determine how the configured QoS rate is distributed to each of the active LAG port SAP queue schedulers, SAP schedulers (H-QoS), and MSS schedulers. These modes are:
adapt-qos distribute
For SAP queue schedulers, SAP schedulers (H-QoS), and SAP egress MSS schedulers, distribute mode divides the QoS rates (as specified by the SLA) equally among the active LAG links (ports). For example, if a SAP queue PIR and CIR are configured on an active/active LAG SAP to be 200 Mb/s and 100 Mb/s respectively, and there are four active LAG ports, the SAP queue on each LAG port will be configured with a PIR of 50 Mb/s (200/4) and a CIR of 25 Mb/s (100/4).
For the SAP ingress MSS scheduler, the scheduler rate is configured on an MDA basis. Distributive adaptive QoS divides the QoS rates (as specified by the SLA) among the active link MDAs proportionally to the number of active links on each MDA.
For example, if an MSS shaper group with an aggregate rate of 200 Mb/s and a CIR of 100 Mb/s is assigned to an active/active LAG SAP where the LAG has two ports on MDA 1 and three ports on MDA 2, the MSS shaper group on MDA 1 will have an aggregate rate of 80 Mb/s (200 ✕ 2/5 of the SLA) and a CIR of 40 Mb/s (100 ✕ 2/5 of the SLA). MDA 2 will have an aggregate rate of 120 Mb/s (200 ✕ 3/5) and a CIR of 60 Mb/s (100 ✕ 3/5).
adapt-qos link (default)
For SAP queue schedulers, SAP schedulers (H-QoS), and SAP egress MSS schedulers, link mode forces the full QoS rates (as specified by the SLA) to be configured on each of the active LAG links. For example, if a SAP queue PIR and CIR are configured on an active/active LAG SAP to be 200 Mb/s and 100 Mb/s respectively, and there are two active LAG ports, the SAP queue on each LAG port will be configured to the full SLA, which is a PIR of 200 Mb/s and a CIR of 100 Mb/s.
For the SAP ingress MSS scheduler, the scheduler rate is configured on an MDA basis. In LAG link mode, each active LAG link MDA MSS shaper scheduler is configured with the full SLA. For example, if an MSS shaper group is configured with an aggregate rate of 200 Mb/s and CIR of 100 Mb/s and is assigned to an active/active LAG SAP with three ports on MDA 1 and two ports on MDA 2, the MSS shaper group on MDA 1 and MDA 2 are each configured with the full SLA of 200 Mb/s for the aggregate rate and 100 Mb/s for the CIR.
Table: Adaptive QoS Rate and Bandwidth Distribution shows examples of rate and bandwidth distributions based on the adapt-qos mode configuration.
|
Distribute |
Link |
|---|---|---|
SAP queue scheduler |
Rate distributed = rate / number of active links |
100% rate configured on each LAG SAP queue |
SAP scheduler (H-QoS) |
Rate distributed = rate / number of active links |
100% rate configured on each SAP scheduler |
SAP egress MSS scheduler |
Rate distributed = rate / number of active links |
100% rate configured on each port’s MSS scheduler |
SAP ingress MSS scheduler |
Rate distributed per active LAG MDA = rate ✕ (number of active links on MDA / total number of active links) |
100% rate configured on each active LAG MDA MSS scheduler |
The following restrictions apply to ingress MSS LAG adaptive QoS (distribute mode).
A unique MSS shaper group must be used per LAG when a non-default ingress MSS shaper group is assigned to a LAG SAP using adaptive QoS.
When a shaper group is assigned to a LAG SAP using adaptive QoS, all ports in the LAG group must have their MDAs assigned to the same shaper policy.
The following restrictions apply to egress MSS LAG.
The shaper policy for all LAG ports in a LAG must be the same and can only be configured on the primary LAG port member.
The following limitations apply to adaptive QoS (distribute mode).
The QoS rates for an ingress LAG using adaptive QoS are only distributed among the active links when a non-default shaper group is used. If a default shaper group is used, the full QoS rates are configured for each port in the LAG as if link mode is being used.
The QoS rates for an ingress or egress LAG using adaptive QoS will not be distributed among the active links when a user sets the PIR/CIR on a SAP queue, or aggregate rate/CIR on a SAP scheduler or MSS scheduler, to the default values (max and 0).