• Author : Ronish Kalia

 
 

VNI based rate limiting

Description This feature enables policer (using policy-map) on a VTEP to rate limit traffic per VLAN/VNI. The policer can be applied in both input and output directions to rate limit decapsulated and encapsulated VXLAN traffic, respectively. The policers will not be applicable on multicast traffic through the VTEP. Platform compatibility 7050SX3(s) 7050CX3(s) 7300X3(s) 7260(s) 7320(s) 7060SX2(s) 7060CX(s) CCS-720XP(s) Configuration arista(config)#class-map match-any class-vxlan arista(config-cmap-qos-t)#match vlan <overlay-vlan-to be monitored> arista(config)#policy-map policy-vxlan arista(config-pmap-qos-t)#class class-vxlan arista(config-pmap-c-qos-t-t)#police rate <X> mbps burst-size <Y> mbytes arista(config)#interface vxlan 1 arista(config-if-Vx1)#service-policy input policy-vxlan * Applies policer on packets being decapped) * arista(config-if-Vx1)#service-policy output policy-vxlan * Applies policer on the...
Continue reading →

Configure ECN max-mark-probability and weight

Introduction Explicit Congestion Notification (ECN) is an extension to the Internet Protocol and to the Transmission Control Protocol which allows end-to-end notification of network congestion without dropping packets. ECN is an optional feature that is only used when both endpoints support it and are willing to use it. ECN operates over an active queue management algorithm. Congestion in ECN is determined by comparing the avg. queue size with thresholds. The average queue size depends on the previous average as well as the current size of the queue and is calculated using the following formula: avg_queue_size = (old_avg * (1-2^(-weight))) +...
Continue reading →

Counters to track headroom peak usage per port per COS

Introduction Headroom buffers are used to store packets when ingress admission control checks fail to store them in both the dedicated service pool derived from the priority of the packet and the shared service pool. This feature monitors the percentage of peak usage of cells in the PG Headroom on a per-port, per COS basis. Platform compatibility 7050CX3- 7300X3- Configuration This feature is enabled by default and cannot be disabled. Status These values are accessible via APIs which can parse smash counters and provide them to the front-end. Arista’s TerminAttr could be used to access these values and display them in the telemetry dashboard. The path...
Continue reading →

PFC Watchdog Enhancements

Platform capability DCS-7050X/X2/X3 series DCS-7060X/X2/X3 series Supported Schedulers In previous releases, PFC Watchdog supported only queues configured with guaranteed bandwidth. Now, it supports all types of schedulers. Forced recovery of queues The Watchdog supports below mechanisms to recover a stuck queue: Auto Recovery – recover queue(s) after the PFC storm ceases Forced Recovery – recover queue(s) after a fixed duration, irrespective of PFC storm being received The default recovery mode is “auto”. Configuration Forced recovery can be enabled with the below CLI command: switch(config) #priority-flow-control pause watchdog default recovery-time <0.01 - 60.0 seconds> forced Syslog Existing syslog for recovery time has...
Continue reading →

Counters to track headroom peak usage per port per COS

Introduction Headroom buffers are used to store packets when ingress admission control checks fail to store them in both the dedicated service pool, derived from the priority of the packet and the shared service pool.  This feature monitors the percentage of peak usage of cells in the PG Headroom on a per-port, per COS basis.  Platform compatibility T3 Configuration This feature is enabled by default and cannot be disabled. Status These values are accessible via APIs which can parse smash counters and provide it to the front-end. TerminAttr from Arista could be used to access these values and display in telemetry dashboard.The...
Continue reading →

Rx/Tx PFC rates

Introduction Ingress PFC rates: The rate of these frames per-port, per-tx-queue would be stored at the path given below, and the polling will happen at the same rate as the other Strata counters. This rate would be applicable globally across ports and is not configurable by the customer and is 1.5 secs. During each poll, the present value of these counters would be subtracted from the previous such value read and the rates will be calculated by dividing this difference by the actual time difference between the two successive reads, from the hardware. Egress PFC rates: These rates will be stored at the path given below...
Continue reading →

Follow

Get every new post on this blog delivered to your Inbox.

Join other followers: