Skip to main content
  • Place orders quickly and easily
  • View orders and track your shipping status
  • Create and access a list of your products
  • Manage your Dell EMC sites, products, and product-level contacts using Company Administration.

Dell EMC Configuration Guide for the S4048T–ON System 9.14.2.4

PDF

Configuring Weights and ECN for WRED

The WRED congestion avoidance functionality drops packets to prevent buffering resources from being consumed. Traffic is a mixture of various kinds of packets. The rate at which some types of packets arrive might be greater than others. In this case, the space on the buffer and traffic manager (BTM) (ingress or egress) can be consumed by only one or few types of traffic, leaving no space for other types. You can apply a WRED profile to a policy-map so that the specified traffic can be prevented from consuming too much of the BTM resources.

WRED drops packets when the average queue length exceeds the configured threshold value to signify congestion. ECN is a capability that enhances WRED by marking the packets instead of causing WRED to drop them when the threshold value is exceeded. If you configure ECN for WRED, devices employ ECN to mark the packets and reduce the rate of sending packets in a congested network.

In a best-effort network topology, data packets are transmitted in a manner in which latency or throughput is not maintained to be at an effective level. Packets are dropped when the network experiences a large traffic load. This best-effort network deployment is not suitable for applications that are time-sensitive, such as video on demand (VoD) or voice over IP (VoIP) applications. In such cases, you can use ECN in conjunction with WRED to resolve the dropping of packets under congested conditions.

Using ECN, the packets are marked for transmission at a later time after the network recovers from the heavy traffic state to an optimal load. In this manner, enhanced performance and throughput are achieved. Also, the devices can respond to congestion before a queue overflows and packets are dropped, enabling improved queue management.

When a packet reaches the device with ECN enabled for WRED, the average queue size is computed. To measure the average queue size, a weight factor is used. This weight factor is user-configurable. You can use the wred weight number command to configure the weight for the WRED average queue size. The mark probability value is the number of packets dropped when the average queue size reaches the maximum threshold value.

The weight factor is set to zero by default, which causes the same behavior as dropping of packets by WRED during network loads or also called instantaneous ECN marking. In a topology in which congestion of the network varies over time, you can specify a weight to enable a smooth, seamless averaging of packets to handle the sudden overload of packets based on the previous time sampling performed. You can specify the weight parameter for front-end and backplane ports separately in the range of 0 through 15.

You can enable WRED and ECN capabilities per queue for granularity. You can disable these functionality per queue, and you can also specify the minimum and maximum buffer thresholds for each color-coding of the packets. You can configure maximum drop rate percentage of yellow and green profiles. You can set up these parameters for both front-end and backplane ports.

Global Service Pools With WRED and ECN Settings

Support for global service pools is now available. You can configure global service pools that are shared buffer pools accessed by multiple queues when the minimum guaranteed buffers for the queue are consumed. Two service pools are used– one for loss-based queues and the other for lossless (priority-based flow control (PFC)) queues. You can enable WRED and ECN configuration on the global service-pools.

You can define WRED profiles and weight on each of the global service-pools for both loss-based and lossless (PFC) service- pools. The following events occur when you configure WRED and ECN on global service-pools:

  • If WRED/ECN is enabled on the global service-pool with threshold values and if it is not enabled on the queues, WRED/ECN are not effective based on global service-pool WRED thresholds. The queue on which the traffic is scheduled must contain WRED/ECN settings, which are enabled for WRED, to be valid for that traffic.
  • When WRED is configured on the global service-pool (regardless of whether ECN on global service-pool is configured), and one or more queues have WRED enabled and ECN disabled, WRED is effective for the minimum of the thresholds between the queue threshold and the service-pool threshold.
  • When WRED is configured on the global service-pool (regardless of whether ECN on global service-pool is configured), and one or more queues are enabled with both WRED and ECN, ECN marking takes effect. The packets are ECN marked up to shared- buffer limits as determined by the shared-ratio for that global service-pool.

WRED/ECN configurations for the queues that belong to backplane ports are common to all the backplane ports and cannot be specified separately for each backplane port granularity. This behavior occurs to prevent system-level complexities in enabling this support for backplane ports. Also, WRED/ECN is not supported for multicast packets.

The following table describes the WRED and ECN operations that occur for various scenarios of WRED and ECN configuration on the queue and service pool. (X denotes not-applicable in the table, 1 indicates that the setting is enabled, 0 represents a disabled setting. )

Table 1. Scenarios of WRED and ECN Configuration Scenarios of WRED and ECN Configuration

Queue Configuration

Service-Pool Configuration

WRED Threshold Relationship

Q threshold = Q-T,

Service pool threshold = SP-T

Expected Functionality

WRED

ECN

WRED

ECN

0

0

X

X

X

WRED/ECN not applicable

1

0

0

X

X

Queue based WRED,

No ECN marking

1

X

Q-T < SP-T

SP-T < Q-T

SP based WRED,

No ECN marking

1

1

0

X

X

Queue-based ECN marking above queue threshold.

ECN marking to shared buffer limits of the service-pool and then packets are tail dropped.

1

X

Q-T < SP-T

SP-T < Q-T

Same as above but ECN marking starts above SP-T.


Rate this content

Accurate
Useful
Easy to understand
Was this article helpful?
0/3000 characters
  Please provide ratings (1-5 stars).
  Please provide ratings (1-5 stars).
  Please provide ratings (1-5 stars).
  Please select whether the article was helpful or not.
  Comments cannot contain these special characters: <>()\