dscp_transparency_test - openconfig/featureprofiles GitHub Wiki

DP-1.17: DSCP Transparency with ECN

Summary

This test evaluates if all 64 combination of DSCP bits are transparently handled while ECN bits are rewritten.

Testbed type

  • TESTBED_DUT_ATE_4LINKS

Procedure

Testbed configuration

  • Connect DUTPort1 with OTGPort1, DUTPort2 with OTGPort2, DUTPort2 with OTGPort2; Assigne IPv4 and IPv6 addresses on all.
  • All 3 ports are of the same speed (100GE)
  • Configure QoS
    • DSCP classifier for IPv4 and IPv4 as below:
      DSCP (dec) Traffic-group
      48-63 NC1
      32-47 AF4
      24-31 AF3
      16-23 AF2
      8-15 AF1
      4-7 BE0
      0-3 BE1
    • 7 queues and 7 corresponding forwarding group
    • Scheduler policy with
      • one scheduler of STRICT priority type serving NC1 queue
      • one scheduler of WRR type serving 6 queues AF4, AF3, AF2, AF1, BE0, BE1 with equal weights 10:10:10:10:10:10 respectively
    • queue-management profile of WRED type with:
      • min-threshold: 80KB
      • max-threshold: 3MB
      • max-drop-percentage: 100
      • ecn: enabled
    • attach queue-management profile to queues NC1, AF4, AF3, AF2, AF1, BE0, BE1;
    • attach scheduler-map to DUTPort3 egress
    • attach classifier to DUTPort1 nad DUTPort2 ingress

Sub Test #1 - No-Congestion

  • Generate 64 flows of traffic form ATEPort1 toward ATEPort3
    • each flow has distinct DSCP value
    • every packet has ECT(0) set
    • all flows of equal bps rate.
    • total load - 60% (60Gbps)
  • wait 1 minutes; stop traffic generation.
  • Verify using DUTPort3 telemetry that:
    • no drops are seen in any of queues on DUTPort3
    • all queues reports non-zero transmit packets, octets.
  • Verify on ATEPort3 that all flows are received w/o DSCP modification -all 64 values are observed
  • verify on ATEPort3 that all received packet has ECT(0) ECN value

Sub Test #2 - Congestion

  • Generate 64 flows of traffic form ATEPort1 and 64 flows of traffic form ATEPort2 toward ATEPort3
    • each flow form ATEPort1 has distinct DSCP value
    • each flow form ATEPort2 has distinct DSCP value
    • every packet has ECT(0) set
    • all flows are of equal bps rate.
    • Offered load:
      • ATEPort1 - 60% (60Gbps)
      • ATEPort2 - 60% (60Gbps)
    • Note: egress port is congested, so do all queues but NC1 (SP)
  • wait 1 minutes; stop traffic generation.
  • Verify using DUTPort3 telemetry that:
    • Drops are seen in all queues except NC1 on DUTPort3
    • all queues reports non-zero transmit packets, octets.
  • Verify on ATEPort3 that all flows are received w/o DSCP modification - all 64 values are observed
  • verify on ATEPort3 that:
    • all received packets with DSCP 48-63 has ECT(0) value
    • vast majority (almost all) packets with DSCP 0-47 has CE ECN value.

Sub Test #3 - NC1 congestion

  • Generate 16 flows of traffic form ATEPort1 and 16 flows of traffic form ATEPort2 toward ATEPort3
    • each flow form ATEPort1 has distinct DSCP value from 48-63 range
    • each flow form ATEPort2 has distinct DSCP value from 48-63 range
    • every packet has ECT(0) set
    • all flows are of equal bps rate.
    • Offered load:
    • ATEPort1 - 60% (60Gbps)
    • ATEPort2 - 60% (60Gbps)
    • Note: egress port is congested, so do NC1 (SP) queue
  • wait 1 minutes; stop traffic generation.
  • Verify using DUTPort3 telemetry that:
    • Drops are seen in NC1 queue on DUTPort3
    • all queues but NC1 reports zero transmit packets, octets.
    • NC1 queue reports non-zero transmit packets, octets.
  • Verify on ATEPort3 that all flows are received w/o DSCP modification - all 16 values are observed.
  • verify on ATEPort3 that:
    • all received packets with DSCP has CE value

OpenConfig Path and RPC Coverage

The below yaml defines the OC paths intended to be covered by this test. OC paths used for test setup are not listed here.

paths:
  ## Config Paths ##
  /qos/classifiers/classifier/terms/term/conditions/ipv4/config/dscp-set:
  /qos/classifiers/classifier/terms/term/conditions/ipv6/config/dscp-set:
  /qos/classifiers/classifier/terms/term/actions/config/target-group:
  /qos/queues/queue/config/name:
  /qos/forwarding-groups/forwarding-group/config/name:
  /qos/forwarding-groups/forwarding-group/config/output-queue:
  /qos/scheduler-policies/scheduler-policy/schedulers/scheduler/config/priority:
  /qos/scheduler-policies/scheduler-policy/schedulers/scheduler/config/sequence:
  /qos/scheduler-policies/scheduler-policy/schedulers/scheduler/inputs/input/config/id:
  /qos/scheduler-policies/scheduler-policy/schedulers/scheduler/inputs/input/config/input-type:
  /qos/scheduler-policies/scheduler-policy/schedulers/scheduler/inputs/input/config/queue:
  /qos/scheduler-policies/scheduler-policy/schedulers/scheduler/inputs/input/config/weight:
  /qos/queue-management-profiles/queue-management-profile/wred/uniform/config/enable-ecn:
  /qos/queue-management-profiles/queue-management-profile/wred/uniform/config/max-drop-probability-percent:
  /qos/queue-management-profiles/queue-management-profile/wred/uniform/config/max-threshold:
  /qos/queue-management-profiles/queue-management-profile/wred/uniform/config/min-threshold:
  /qos/interfaces/interface/output/queues/queue/config/name:
  /qos/interfaces/interface/output/queues/queue/config/queue-management-profile:
  /qos/interfaces/interface/output/scheduler-policy/config/name:
  /qos/interfaces/interface/input/classifiers/classifier/config/name:
  /qos/interfaces/interface/input/classifiers/classifier/config/type:
    
  ## State Paths ##
  /qos/interfaces/interface/output/queues/queue/state/dropped-octets:
  /qos/interfaces/interface/output/queues/queue/state/dropped-pkts:
  /qos/interfaces/interface/output/queues/queue/state/name:
  /qos/interfaces/interface/output/queues/queue/state/transmit-octets:
  /qos/interfaces/interface/output/queues/queue/state/transmit-pkts:

rpcs:
  gnmi:
    gNMI.Set:
    gNMI.Subscribe:

Required DUT platform

  • FFF
⚠️ **GitHub.com Fallback** ⚠️