Bug 2137597

Summary:	Performance degradation for OVS HWOL RHEL 8.6 on ConnectX6-Lx and BF-2 (ConnectX mode)
Product:	Red Hat Enterprise Linux Fast Datapath	Reporter:	Gurpreet Singh <gurpsing>
Component:	openvswitch	Assignee:	Marcelo Ricardo Leitner <mleitner>
openvswitch sub component:	ovs-hw-offload	QA Contact:	ovs-qe
Status:	NEW ---	Docs Contact:
Severity:	high
Priority:	high	CC:	ctrautma, qding, rkhan
Version:	RHEL 8.0
Target Milestone:	---
Target Release:	---
Hardware:	Unspecified
OS:	Unspecified
Whiteboard:
Fixed In Version:		Doc Type:	If docs needed, set a value
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:		Type:	Bug
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:
Bug Depends On:
Bug Blocks:	2172622

Description Gurpreet Singh 2022-10-25 15:30:24 UTC

Description of problem:

Severe degradation of throughput is observed for ConnectX6-LX and BF-2 (ConnectX mode) with RHEL 8.6 OVS HWOL as compared to throughput on ConnectX-5 and ConnectX6-Dx NICs. Following is the link to the test results. The performance degradation is more severe for 10k flows as compared to 1k flows.

https://docs.google.com/spreadsheets/d/1j6dHdj3V47UJyXL5bm4zoV349ia1ZCzDXX-dRwvSc2E/edit#gid=653599643



Version-Release number of selected component (if applicable):


How reproducible:


Steps to Reproduce:
1.
2.
3.

Actual results:


Expected results:


Additional info:

Comment 2 Marcelo Ricardo Leitner 2022-12-29 15:56:03 UTC

This needs to be discussed with Nvidia. I don't see any reasons for this to happen, other than possibly different FW configs.
Can we share this spreadsheet with them as well? I can add the relevant people if needed once I get an ack.

Comment 4 Marcelo Ricardo Leitner 2023-02-23 03:58:06 UTC

I have the impression that this shares a root cause with https://bugzilla.redhat.com/show_bug.cgi?id=2143570
10k flows can put a toll on it.