Administrator Guide

Guidelines for Monitoring High-Gigabit Port Channels
Keep the following points in mind when you activate and examine the utilization and working-efficiency
of backplane high-Gigabit Ethernet port channels as trunk groups:
• By default, the capability to monitor the traffic utilization and distribution of high-Gigabit Ethernet
trunk groups is disabled.
• Each NPU unit in each line card (or control processor card) can contain multiple trunk groups (high-
Gigabit port channels). The interface specifier convention for hg-port-channel is slotId/npuUnitId/
localPortChannelId, which denotes the slot, NPU, and the port channel identifiers.
• For Z9000, slotId (stack unitId) is constant and does not vary. NpuUnitId ranges from 0-5 and local
portChannelId ranges from 0-0 for leaf NpuUnits and 0-3 for spine NpuUnits.
• Link-bundle monitoring starts if monitoring is enabled for the bundle whose egress utilization exceeds
a threshold. This behavior is required if you want to view the utilization alarms only when the
utilization levels are high. At low utilization levels, it is possible that there are only one or two
significant flows that are causing unevenness. Such an imbalanced traffic flow is not critical or
indicative of a problem. The hiGig link bundle trigger threshold is a system-wide or a global setting for
the device.
• If you enabled the generation of SNMP traps, syslogs and traps are transmitted when an uneven
distribution is observed. Another syslog and trap is generated when the unevenness is cleared.
• Link bundle utilization is calculated as the bandwidth-weighted mean utilization of all links in a bundle
(calculated as [total bandwidth of all links / total bytes-per-sec of all links]). This calculation is
performed only on those links that are up on their operational status.
• The rate interval for polling the traffic statistics for member links of the high-Gigabit port channel
must be configured. The default hiGig stats polling interval is 15 seconds. This interval cannot be
configured per high-Gigabit port channel and is applicable for all of the high-Gigabit port channels on
the system.
• The threshold value identifies when to start the link bundle utilization calculation trigger (default of 60
percent). When overall utilization (mean) is below this value, link bundle distribution unevenness is not
reported.
• If unevenness is observed over three consecutive measurements, an alarm event is generated. The
rateInterval for Hg stats polling (default 15 seconds) determines the time interval between two
measurements. Alarm clear is sent when evenness is observed for three successive rate intervals. If
individual link utilization information is not available for a given timestamp, link bundle utilization is
not calculated at that timestamp. The previous known record shall be used for the alarm calculation.
• Turning on and off the link bundle monitoring is performed at a high-Gigabit Ethernet port-channel
level configuration.
• The difference of utilization % between the high-used link and low-used link is used to determine the
alarm condition. The alarm trigger reporting is based on the same algorithm used for link bundle
monitoring on LAG/ECMP. The alarm reporting is triggered when the configured threshold has
crossed for a given bundle. Now, if the delta utilization is beyond 10%, an alarm is raised. The alarm
condition remains active until all interface utilizations are within the band or until the overall utilization
goes below the trigger threshold. An alarm is not raised or cleared instantaneously
Link Aggregation Control Protocol (LACP)
441