Re: RFC: Dragonfly


AllenSun
 

Hi, Deepak Vij,

Thanks for your feedback.
Yes, it's a great job to support image pulling optimization in scheduling of Kubernetes. 

I think that the important feature you mentioned is 

  • focus on scheduler's scheduling algorithm;
  • image layer tracking information of node is a factor of scheduler;
  • scheduling algorithm may reduce the image pulling if scheduling workload to a node if image layer exists on that node.

So, image pulling optimization you mentioned in on scheduling, while Dragonfly is mainly on Image Distribution. I think they are decoupled and only have few of overlapping.
In addition, I think this feature and Dragonfly can co-exist very well to help improve image pulling. The feature decides the workload location by scheduling, and Dragonfly can improve the rest image distribution.

Dragonfly takes over the image distribution for Kubernetes, and image pulling efficiency is only one part of distribution. Besides this, it also provides:

  • image distribution flow control;
  • image distribution security.

Currently all the three parts above consist of Dragonfly's service.

Best Regards.

Allen Sun
Alibaba Group



------------------------------------------------------------------
Sender:Deepak Vij <deepak.vij@...>
Sent at:2018 Sep 5 (Wed) 09:20
Recipient:Allen <allensun.shl@...>; cncf-toc <cncf-toc@...>
Subject:Re: [cncf-toc] RFC: Dragonfly

While we are on the topic of image pulling optimizations, following granular image layers level optimizations are being worked on as part of Kubernetes Scheduling SIG future release feature:  

1)      Layer tracking system on each machine/node discovers the layer information from local file-system and reports this information to workload scheduler (for example, K8S scheduler in our case).

2)      As part of the image resolution process for an incoming container workload, image request is decomposed to underlying layer requests.

3)      Scheduler uses these two pieces of information to perform workload scheduling in accordance to the layering information on the node.

Regards,

Deepak Vij

 

From: cncf-toc@... [mailto:cncf-toc@...] On Behalf Of AllenSun
Sent: Tuesday, September 04, 2018 10:56 AM
To: cncf-toc <cncf-toc@...>
Subject: Re: [cncf-toc] RFC: Dragonfly

 

 

Hi, Friends,

 

It is my honor to give a presentation about Dragonfly to CNCF community. Dragonfly aims at covering all the image distribution in CNCF ecosystem. And currently it works perfectly within Kubernetes. My slides are in the attachment of this email.

 

Please feel free to give us your feedback on Dragonfly. Any thoughts on it are welcome.

 

In addition, I am really happy that Jonathan Boulle would like to sponsor Dragonfly to enter sandbox level.

While we sincerely wish that an additional TOC member could sponsor this.

 

How to get help from Dragonfly's team, if you are interested in this, please don't hesitate to drop questions on me.

 

Thanks a lot.

 

Allen Sun

Alibaba Group

 

 

 

 

------------------------------------------------------------------

Sender:Chris Aniszczyk <caniszczyk@...>

Sent at:2018 Sep 4 (Tue) 23:33

Recipient:CNCF TOC <cncf-toc@...>

Cc:Allen <allensun.shl@...>

Subject:[cncf-toc] RFC: Dragonfly

 

The Dragonfly project was presented to the TOC today:

 

They currently have one TOC sponsor, Jonathan Boulle.

 

The Dragonfly community is seeking an additional TOC sponsor and happy to answer any questions from the community.

 

--

Chris Aniszczyk (@cra) | +1-512-961-6719

 


Join cncf-toc@lists.cncf.io to automatically receive all group messages.