Yes, agreed, more proposals and participation are always good.
In terms of the lab models that George asked about, we decided early in our discussions to implement a distributed CI solution. Thomas has done great work to
integrate this with Patchwork, so you can see which CI tests each patch has passed/failed at:
http://dpdk.org/dev/patchwork/project/dpdk/list/ (see the Success/Warning/Fail (S/W/F) columns).
There was interest in creating a community lab to be hosted by the Linux Foundation for independent performance testing and for identifying performance regressions.
It was felt that results from an independent lab would have more credibility than results from vendor labs. As Heqing said, we’ll have a proposal on this after the PRC New Year holiday. If we end up not having budget for this, or don’t agree that it’s required,
then we can explore other options such as a distributed solution.
From: Zhu, Heqing
Sent: Wednesday, January 25, 2017 5:39 PM
To: George Zhao <George.Y.Zhao@huawei.com>; O'Driscoll, Tim <tim.odriscoll@intel.com>; Ed Warnicke <hagbard@gmail.com>
Cc: Wiles, Keith <keith.wiles@intel.com>; moving@dpdk.org; Zhu, Heqing <heqing.zhu@intel.com>
Subject: RE: [dpdk-moving] Minutes from "Moving DPDK to Linux Foundation" call, January 24th
As matter of fact today, Intel helps the DPDK release validation. As Tim said, we will make a proposal after CNY.
DPDK is an open community, more proposals and participation are welcomed.
From: moving [mailto:moving-bounces@dpdk.org]
On Behalf Of George Zhao
Sent: Wednesday, January 25, 2017 9:29 AM
To: O'Driscoll, Tim <tim.odriscoll@intel.com>; Ed Warnicke <hagbard@gmail.com>
Cc: Wiles, Keith <keith.wiles@intel.com>;
moving@dpdk.org
Subject: Re: [dpdk-moving] Minutes from "Moving DPDK to Linux Foundation" call, January 24th
I know there are two ways normally community lab operate, one is like fd.io project, CSIT lab is managed by Linux Foundation, the
other is like OpenDaylight where member companies open their lab to share with community.
Do we decide which way for DPDK lab?
George
From: moving [mailto:moving-bounces@dpdk.org]
On Behalf Of O'Driscoll, Tim
Sent: Wednesday, January 25, 2017 9:10 AM
To: Ed Warnicke
Cc: Wiles, Keith; moving@dpdk.org
Subject: Re: [dpdk-moving] Minutes from "Moving DPDK to Linux Foundation" call, January 24th
We haven’t yet agreed that we definitely need a lab, how big it needs to be, and how much it will cost. Our team in PRC have been working
on a proposal, but with their New Year holidays that’s a couple of weeks away from being ready to share with the community.
The scope we’ve been discussing for the lab is quite small when compared to FD.io’s CSIT project. It would be a reference lab to provide
independent performance data and to identify any performance regression. The ~$200k I quoted for a single rack is really the minimum starting point. If we agree we need more and have the budget to cover it, then we can expand beyond that.
Mike will explore interest in the lab as part of his discussions, and we’ll also have one of our PRC team present the proposal to the
community when they return from their New Year holiday. After that, we’ll know more about the level of interest in the lab and the cost associated with it.
Tim
From: Ed Warnicke [mailto:hagbard@gmail.com]
Sent: Wednesday, January 25, 2017 4:31 PM
To: O'Driscoll, Tim <tim.odriscoll@intel.com>
Cc: Wiles, Keith <keith.wiles@intel.com>;
moving@dpdk.org
Subject: Re: [dpdk-moving] Minutes from "Moving DPDK to Linux Foundation" call, January 24th
Question... are you only pricing for *one* rack? I ask, because *one* rack can fill pretty quick...
Ed
On Wed, Jan 25, 2017 at 9:16 AM, O'Driscoll, Tim <tim.odriscoll@intel.com> wrote:
> From: Wiles, Keith
>
> Sent from my iPhone
>
> > On Jan 25, 2017, at 4:57 AM, O'Driscoll, Tim <tim.odriscoll@intel.com>
> wrote:
> >
> >
> > Membership costs:
> > - Discussed potential membership costs. My proposal was ~$50-100k for
> Gold, ~$5-$20k for Silver. Most agreed that this was a good starting
> point for discussions.
>
> I thought we were trending toward the higher $100k range as the lab was
> going cost a fair bit am I wrong here?
The membership rates we decide on will need to strike a balance between raising budget and having a broad membership that's representative of the breadth of DPDK contributions/usage. If we choose a high figure it will limit the number of companies prepared to join. If we choose too low a number then we won't maximize our budget. We need to strike a balance between the two.
The next step we agreed was for Mike to identify who's interested in membership (he's already posted on the moving list asking for contacts) and begin to have individual discussions with them. Feedback on membership rates from these discussions will help us to make a final decision.
I think we need to be careful on lab costs. Some high figures have been mentioned based on FD.io, but from the beginning of these discussions we've agreed that we want a smaller scope and lower cost level for DPDK. Rough estimate for a full rack with a part time sys admin and a part time release engineer is ~$200k/year.
We also discussed yesterday whether lab costs should be fully accounted for in the Gold membership fee, or if they should be handled separately. Mike will also ask about interest in the lab as part of his discussions which will help us to reach a conclusion on this.