The 7705 SAR supports single-ended and dual-ended Ethernet Loss Measurement (ETH-LM) tests. Single-ended LM tests are run using the oam>eth-cfm>single-ended-loss-test command and are considered on-demand tests. Dual-ended LM tests are enabled using the eth-cfm>mep>dual-ended-loss-test-enable command for Epipe services and network router interfaces. When enabled, dual-ended LM tests run continuously in the background.
Dual-ended LM tests are not supported for VPLS.
Y.1731 loss measurement functionality is implemented to ensure the most accurate results under all circumstances. Each adapter card has a network processor (NP). LM counters are maintained at the NP. The NP is responsible for incrementing and resetting these counters. These counters are accessed by the CSM CPU in order to calculate and display the loss (percentage) to the user.
LM/CCM frames follow the associated QoS path and therefore might inadvertently report loss due to local congestion even before the frame is switched onto the link. In order to reflect the true experience of a particular QoS setting, generated LM/CCM frames follow the egress QoS path. Once generated, these frames are classified in the same manner as the applicable dot1p-to-FC mapping, associated queuing, and scheduling rules. Following the proper path ensures that loss measurements reflect the experience of a given FC all the way through the network, including within the 7705 SAR platform. As is the case for any other frame of the same FC (that is, user or control frame), the LM/CCM frame follows the associated QoS path to reflect the real experience.
For example, newly generated LM/CCM frames that have a higher counter value can be forwarded sooner than LM/CCM frames with a lower counter value that have been generated but are waiting to be serviced (that is, frames with a lower queue, a queue in the out-of-profile state; or a single SAP with multiple FCs). As a result, when under congestion, the LM ratio would increase to reflect local loss if lower-priority frames cannot be serviced in a timely manner.
In addition, congestion, and hence prioritization, can occur anywhere in the transport network, which means that a reordering could take place not only on the ingress point, but anywhere in the network along the entire path.
The loss ratio is calculated based on the aggregate frames being transmitted and received. In an uncongested network, the loss ratio would be 0%. With congestion, not all frames may be sent out to the network (that is, higher priority traffic, and so on) or any one of the transit nodes or the endpoint node might drop the packet, which would end up with loss.
The above-described behavior for following the QoS path equally applies to both Up and Down MEPs. Loss measurements in both up and down directions for the same MEP can be performed simultaneously.
The counters used for loss measurement in LM and CCM frames are appended as late as possible in the datapath. Appending the counters at the last minute to the LM or CCM frames ensures that a scheduling priority issue or some other queue-delaying event does not delay the OAM frame in a queue. If the counters are updated or generated earlier in the datapath, then the OAM frames could be affected by queuing or scheduling delays, which could cause the frames to be counted as lost frames when the far-end receive timer expires.
The following notes apply to Y.1731 LM tests.
Single-ended and dual-ended LM tests cannot be enabled on a MEP simultaneously. That is, either a single-ended or a dual-ended LM test can be enabled on a specific MEP at any one time.
The behavior and the interaction between single- and dual-ended LM tests are described in the following list. Error conditions, such as correct domain level and valid destination address (DA) MAC, are not covered in the list:
if dual-ended LM test is disabled:
CCM frames are transmitted with LM counters set to 0
CCM frames being received are not processed for LM
LMM and LMR frames being received are processed
single-ended tests can be enabled (not blocked by CLI)
if dual-ended LM test is enabled:
single-ended tests cannot be enabled (blocked by CLI)
LMM and LMR frames being received will be dropped
Multiple MEPs bound to the same Epipe SAP but belonging to different MEG levels can perform LM tests simultaneously.
CCM must be enabled before a dual-ended LM test can be enabled.
When a dual-ended LM test is enabled, the user cannot disable CCM. The dual-ended LM test needs to be disabled before the CCM can be disabled.
For dual-ended LM tests, an alarm is declared when frame losses are greater than an alarm threshold configured for the MEP. The granularity of the alarm threshold (declaring or clearing) is 0.01%. The default threshold is set to 0.25%.
On a per-SAP or per-network interface basis, there is one set of Rx and Tx Local LM counters and one set of Rx and Tx Remote LM counters. LM counters are not separated on a per-MAC source address (SA) basis. All MEPs, regardless of their MD or MEG level, share the same set of Rx and Tx LM counters.
On the CLI, there are interval counters and accumulated counters. The CCM counters are referred to as Local and FarEnd counters and the accumulated counters are referred to as Near-End and Far-End counters.
The LM counters are incremented when a user data frame reaches a SAP. Because there is only one set of Tx and Rx Local counters per SAP, each user data frame received by all the MEPs configured on that SAP is counted.
OAM frames with MEP levels matching or lower than the locally configured MEP level are not counted. They are treated and processed as OAM frames. This functionality applies to both received and transmitted OAM frames. CFM OAM frames at higher MEP levels are counted as user data frames.
For example, assume a SAP with two MEPs configured on it; one MEP at level 5 and the other at level 6.
When a level 6 OAM frame is received, it is extracted to the CSM for processing and is not counted by LM counters. It is treated as an OAM frame.
The same behavior applies in the transmit direction. In the above example, any level 5 or level 6 OAM frames generated by the local SAR would not be counted by the far-end LM counters.
For dual-ended LM tests, any received CCMs with all LM counters being 0s (zeros) are treated as invalid. In this case, the 7705 SAR resets the LM counters for the current and previous CCMs to 0s (zeros). Accumulated counters are not reset.
Except for a valid counter rollover scenario, if the value of any CCM/LMR counter is less than the value of the same counter in the previous CCM/LMR frame, then the accumulated values of all counters are not increased; they are kept at the same values as before the last CCM/LMR frame is processed.
When the first valid CCM/LMR frame — that is, a frame with at least one non-zero LM counter — is received after a dual-ended loss test is enabled or a single-ended loss test is launched, the accumulated values cannot be calculated. In this case, the counters are resaved as current counters. When the next received CCM/LMR frame with valid LM counts is received, it will trigger the update of accumulated counts.
Accumulated counts always start at 0 for each launch of a single-ended test. However, the accumulated counts do not change nor do they get reset to all 0s when dual-ended loss tests become disabled. For dual-ended loss tests, accumulated counts can be restarted at 0 by removing the existing LM result of a particular MEP with the CLI command clear>eth-cfm>dual-ended-loss-test>mep mep-id domain md-index association ma-index, or the equivalent SNMP command.