Active Buffer Monitoring Overview
The Active Buffer Monitoring feature provides detailed buffer occupancy data to help you detect network congestion, review past events to understand when and how network congestion is affecting network operations, understand historical trending, and identify patterns of application traffic flow.
A hardware component, called the Algorithm Boost Engine (Algo Boost Engine) supports buffer histogram counters for unicast buffer usage per individual port, total buffer usage per buffer block, and multicast buffer usage per buffer block. Each histogram counter has 18 buckets that span across the memory block. The Algo Boost Engine polls buffer usage data every hardware sampling interval (the default is every 4 milliseconds, but you can configure it to be as low as 10 nanoseconds). Based on the buffer utilization, the corresponding histogram counter is incremented. For example, if Ethernet port 1/4 is consuming 500 KB of the buffer, the bucket 2 counter (which represents 384 KB to 768 KB) for Ethernet 1/4 is incremented.
To avoid a counter overflow, the Cisco NX-OS software collects the histogram data every polling interval and maintains it in the system memory. The software maintains the histogram data in the system memory for the last 60 minutes with 1-second granularity. Every hour, the software copies the buffer histogram data from the system memory to the bootflash as a backup.
The Active Buffer Monitoring feature has two modes of operation:
Unicast mode—The Algo Boost Engine monitors and maintains a buffer histogram for total buffer utilization per buffer block and unicast buffer utilization for all 48 ports.
Multicast mode—The Algo Boost Engine monitors and maintains buffer histogram data for total buffer utilization per buffer block and multicast buffer utilization per buffer block.