Minutes of Techboard Meeting, 2024-01-10

Minutes of Techboard Meeting, 2024-01-10

[Honnappa Nagarahalli]

TB Attendees
-----------------
Bruce
Hemant
Honnappa
Jerin
Kevin
Konstantin
Maxime
Morten
Thomas


NOTE: The technical board meetings are on every second Wednesday at 3 pm UTC. Meetings are public, and DPDK community members are welcome to attend.
Link to join:
https://zoom-lfx.platform.linuxfoundation.org/meeting/96459488340?password=d808f1f6-0a28-4165-929e-5a5bcae7efeb
NOTE: Next meeting will be on Wednesday 2024-Jan-24 @3pm UTC, and will be chaired by Kevin

1. Governing board direction to Techboard
- Governing board has asked Techboard to explore the applicability and usability of DPDK in certain technology areas and address any gaps.
- This seems like an expansion of the role of Techboard. Originally Techboard was created as a final arbiter of the patches coming from DPDK community.
- This will add burden on Techboard members' bandwidth.
- This is a community driven project and not a Techboard driven project.
- If the investigations find gaps in the technology areas, the Techboard still expects contributions from member companies to address these gaps.
- If the technology areas are of interest to member companies, Techboard expects them to contribute patches to the community.
- All that the Techboard could do is, drive attention to the gaps.
- Set realistic expectations on the outcome given constraints on bandwidth, knowledge and lack of contributions
- Thomas to understand the role of DPDK in AI

2. TCP and TLS support in DPDK
- Does TLS slow path need support in DPDK?
- There are performance and ease of use benefits for creating a DPDK native TCP stack.
- We need to think of adding UDP support which is being used more widely. In general we need to identify gaps to support higher level protocols.

3. How well does DPDK run on Hyperscaler environments?
- Can UNH take the task of running DPDK on Hyperscaler environments? UNH seems to have lot of things already.
- In the Hyperscaler environment, DPDK could be used for various purposes. DPDK usage in Hyperscaler infrastructure (for ex: DPU), the existing DTS testing could be sufficient.
- However, the DPDK usage in tenant deployment scenarios needs to be tested.
- Could cloud providers send their performance and functional test reports to DPDK community (similar to how we publish these reports for physical NICs)?
- We can start with Gov board to place these requests to members on Gov board.
- We could also ask the maintainers of the virtual NICs from the cloud providers.
- Thomas to follow up on the above two asks.

4. The security topic would be discussed in a joint TB and GB meeting scheduled for next Thursday (2024-Jan-18).

5. Pending EAL arg parse patches
- There are 3 separate patches on argument handling for DPDK.
	a. Python argparse equivalent - http://patches.dpdk.org/project/dpdk/list/?series=30506
	b. library to build up a argv, argc list to pass it to EAL init. This is not a complex problem to solve. But it is helpful when one has a more dynamic list of arguments - http://patches.dpdk.org/project/dpdk/list/?series=30120
	c. Simplest and smallest. Provides core mask and core list parsing - https://patches.dpdk.org/project/dpdk/list/?series=30582
- These are all slightly different.
- Ask for Techboard: Go through the patches, participate in discussions on the mailing list.
- Need to decide the functionality required and decide on which way to move forward.
- Does it make sense to consider if the library can also be used for managing the arguments for Telemetry?
- Investigate the ability to initialize/create things through APIs, not only though EAL text arguments via argc/argv.

Re: Minutes of Techboard Meeting, 2024-01-10

[Patrick Robb]

[-- Attachment #1: Type: text/plain, Size: 1502 bytes --]

On Tue, Jan 16, 2024 at 11:45 AM Honnappa Nagarahalli <
Honnappa.Nagarahalli@arm.com> wrote:

>
> 3. How well does DPDK run on Hyperscaler environments?
> - Can UNH take the task of running DPDK on Hyperscaler environments? UNH
> seems to have lot of things already.
>

Sorry, I could not make techboard this week as I had a conflict from
another project. I would like to reiterate that in principle an effective
way to start addressing any gaps is to use testing to better understand the
landscape as is. At the same time, yes this conversation did not reach
maturity during 2023, so there aren't hyperscaler testing work items in
2024's Community Lab SOW. Written SOW items always have to take precedence,
but we will continue to accept ad-hoc community requests like always, and
engage in conversations for what's next for testing in DPDK.

We had an email chain going with Robin/Honnappa about the most needed test
coverage in this space, etc. but it died out. I remember Honnappa
mentioning the importance of measuring multi core scalability performance,
running l2fwd l3fwd sample apps etc. I think Robin may host some calls in
2024 to take feedback on this. We would still need a plan on funding for
the cloud instances used, list of vNICs to provide testing coverage for, a
test plan, etc. So, realistically today we are very far off, but if the
conversation on this continues to develop, we can engage and define better
if/how the Community Lab could play a role.

[-- Attachment #2: Type: text/html, Size: 1886 bytes --]