Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[Broadcom-DNX][Nokia] ECN marking is not happening at configured Kmin and Kmax value. #20146

Open
amitpawar12 opened this issue Sep 5, 2024 · 2 comments
Labels
Chassis 🤖 Modular chassis support

Comments

@amitpawar12
Copy link

Description

This issue is reported in response to the ECN test-cases present in snappi_tests/multidut/ecn folder.

Steps to reproduce the issue:

  • Configured Kmin and Kmax value to 800k and 1200k bytes.
  • Set Pmax to 100.
  • Configure alpha threshold to 3 (2^3) [This step is optional and does not alter testcase functionality].
  • Send pause on the egress port.
  • Send 5k packets with 1024B.

Describe the results you received:

  • All packets from 267 to around 4996 are ECN-marked.
  • This does not change even if the Kmin and Kmax thresholds are changed.

Describe the results you expected:

We have a Broadcom CSP open to understand the queuing/de-queuing aspects involved in ECN marking.

Output of show version:

(paste your output here)

Output of show techsupport:

(paste your output here or download and attach the file here )

Additional information you deem important (e.g. issue happens only occasionally):

Tagging @saksarav-nokia

@saksarav-nokia
Copy link
Contributor

Following with BCM with CS00012362297

@zhangyanzhao zhangyanzhao added the Chassis 🤖 Modular chassis support label Sep 11, 2024
@saksarav-nokia
Copy link
Contributor

BCM's comments:
When unicast packets are received on a NIF port they are then sent to a VOQ based on processing in the ingress pipeline (IRPP). In congestion scenarios these packets are then sent to DRAM when thresholds are crossed.
Also, the packets are marked with ECN on dequeue. This is why you are seeing the packets marked before the min threshold when the queue is heavily congested.
The CNI status is tagged per DRAM bundle instead of per packet when the packets are sent from SRAM to DRAM. This is why you were also seeing most of the last packets being marked when they were going to DRAM. This is not the case when all of the packets go to OCB.
The ECN test with XGS or other pizza boxes where we count ECN marked packets is going to be very difficult with DNX with different thresholds (because of the drop probabilities and DRAM bundling).

A simpler approach might be to have a test with the kmin setting where you see no marking and then another test in between kmin and kmax where you want to confirm that packets are being marked.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Chassis 🤖 Modular chassis support
Projects
None yet
Development

No branches or pull requests

3 participants