Arch Committee Election: Q&A period open!
Hello all, We have three candidate submissions for the Architecture Committee election, where we have two seats available. Thus, we shall enter the Q&A and subsequently the voting phases. Candidates names and links to their github submissions are below. The next period (Feb13-23rd) is for asking questions to candidates via this mailing list; it's an opportunity to get familiar with the perspective they would bring as an Arch Committee member before voting opens on the 24th. To kick off the Q&A, I shall ask to the candidates: Q. What is the most important thing (be it an integration, a particular feature, adopting a policy, etc) Kata Containers needs to achieve in the next year and how should we get there? Candidates (in alphabetical order): -David Alan Gilbert [1] -Samuel Ortiz [2] -Xu Wang [3] [1] https://github.com/kata-containers/community/blob/master/elections/arch-comm... [2] https://github.com/kata-containers/community/blob/master/elections/arch-comm... [3] https://github.com/kata-containers/community/blob/master/elections/arch-comm... --------------------------------------------------------------------- Intel Corporation (UK) Limited Registered No. 1134945 (England) Registered Office: Pipers Way, Swindon SN3 1RJ VAT No: 860 2173 47 This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies.
Thank you, Graham.
Q. What is the most important thing (be it an integration, a particular feature, adopting a policy, etc) Kata Containers needs to achieve in the next year and how should we get there?
As a community, we need more participants from end-users. To get there, we should work on three aspects: reducing overhead, improving isolation, and helping new users: More and more people agree with isolation is important, however, they still need to pay too much to migrate to kata containers. I think we should work on the following aspects: - Reducing memory overhead. People like isolation, but they don't want to pay for it. - `rust-agent` is a good starting point, and we could reduce the heap memory consumption down to less than 800KB in our initial tests with rust-agent+ttrpc. - `cloud-hypervisor` has been integrated with kata, and it is a lightweight, modularized, and more open project for contributors. We shall work with it and have a kata profile of it. - Being more flexible. One critique on Kata is that we are not as flexible as a "real" container. - We are working on virtio-mem, @teawater from our team has prepared patches and documents for it, and he is working on virtio-mem for cloud-hypervisor as well. I am looking forward to landing it in kata in months. On the other hand, we should enhance the isolation of Kata, which is our initial goal. - Enable users to move all the image operations in the guest sandbox for better isolation. In a discussion during the last PTG, the developers from Alibaba and IBM are all working on it. - It is possible to introduce TEE technologies from Intel/AMD/IBM/ARM chips, which provide "bi-direction isolation", i.e. protecting workloads from the infra providers. Another important thing is providing more user practice in the community, which may help new adopters. - Baidu's Kata deployment won the Superuser Award in the last summit and has a whitepaper published. We are working on Alibaba and Ant Financial Kata clusters and will publish more document and testing data about them. And we still need more from all the contributors. - The marketing team of Kata is planning on setting up neutral and open performance metrics test for Kata, which will help the users to know whether or when Kata is fit for them. - And we shall attend conferences and meetups on infra and cloud-native to promote Kata as well. Thank you guys for the contributions, and let's work together for a better year. Cheers! Xu ------------------------------------------------------------------ From:Whaley, Graham <graham.whaley@intel.com> Send Time:2020年2月13日(星期四) 18:25 To:kata-dev@lists.katacontainers.io <kata-dev@lists.katacontainers.io> Subject:[kata-dev] Arch Committee Election: Q&A period open! Hello all, We have three candidate submissions for the Architecture Committee election, where we have two seats available. Thus, we shall enter the Q&A and subsequently the voting phases. Candidates names and links to their github submissions are below. The next period (Feb13-23rd) is for asking questions to candidates via this mailing list; it's an opportunity to get familiar with the perspective they would bring as an Arch Committee member before voting opens on the 24th. To kick off the Q&A, I shall ask to the candidates: Q. What is the most important thing (be it an integration, a particular feature, adopting a policy, etc) Kata Containers needs to achieve in the next year and how should we get there? Candidates (in alphabetical order): -David Alan Gilbert [1] -Samuel Ortiz [2] -Xu Wang [3] [1] https://github.com/kata-containers/community/blob/master/elections/arch-comm... [2] https://github.com/kata-containers/community/blob/master/elections/arch-comm... [3] https://github.com/kata-containers/community/blob/master/elections/arch-comm... --------------------------------------------------------------------- Intel Corporation (UK) Limited Registered No. 1134945 (England) Registered Office: Pipers Way, Swindon SN3 1RJ VAT No: 860 2173 47 This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies. _______________________________________________ kata-dev mailing list kata-dev@lists.katacontainers.io http://lists.katacontainers.io/cgi-bin/mailman/listinfo/kata-dev
* Whaley, Graham (graham.whaley@intel.com) wrote:
Hello all,
We have three candidate submissions for the Architecture Committee election, where we have two seats available. Thus, we shall enter the Q&A and subsequently the voting phases. Candidates names and links to their github submissions are below.
The next period (Feb13-23rd) is for asking questions to candidates via this mailing list; it's an opportunity to get familiar with the perspective they would bring as an Arch Committee member before voting opens on the 24th.
To kick off the Q&A, I shall ask to the candidates: Q. What is the most important thing (be it an integration, a particular feature, adopting a policy, etc) Kata Containers needs to achieve in the next year and how should we get there?
I think it's important for Kata to become easy to use by those who are already familiar with containers, and just want to flip on the extra protection. That means we need to provide them some easily understandable idea of the overhead, and make it easy for them to debug problems. These users might be those who have a different focus and aren't necessarily large cloud admins. Dave
Candidates (in alphabetical order): -David Alan Gilbert [1] -Samuel Ortiz [2] -Xu Wang [3]
[1] https://github.com/kata-containers/community/blob/master/elections/arch-comm... [2] https://github.com/kata-containers/community/blob/master/elections/arch-comm... [3] https://github.com/kata-containers/community/blob/master/elections/arch-comm...
--------------------------------------------------------------------- Intel Corporation (UK) Limited Registered No. 1134945 (England) Registered Office: Pipers Way, Swindon SN3 1RJ VAT No: 860 2173 47
This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies.
_______________________________________________ kata-dev mailing list kata-dev@lists.katacontainers.io http://lists.katacontainers.io/cgi-bin/mailman/listinfo/kata-dev -- Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK
Hi All, On Thu, Feb 13, 2020 at 10:25:00AM +0000, Whaley, Graham wrote:
Q. What is the most important thing (be it an integration, a particular feature, adopting a policy, etc) Kata Containers needs to achieve in the next year and how should we get there?
Since early on, Kata Containers threat model has been centered around protecting the host from malicious container workloads. During the next year, I'd like to see that threat model being extended to also protecting container workloads from the host infrastructure itself. I believe the Kata Containers architecture can be expanded to address that rising use case where containers can be securely deployed over untrusted hosts by: * Handling the CRI image operations from the guest itself. This is a fairly intrusive change that will impact many layers of the stack (Runtimes, kata agent, kata runtime) and our performance overhead as well. * Enabling and using hardware based total memory encryption technologies in order to prevent the host infrastructure from reaching the container assigned memory. I would also like to work on making things easier for newcomers to contribute to the project. And for that I'd like to use Kata Containers 2.0 as an opportunity for our community to prune the list of supported features and configuration knobs, and to also reduce our code complexity. Both aspects are related and will not only benefit potential newcomers but also our existing contributors and users. Cheers, Samuel.
Hello Kata Citizens! We have four candidate submissions for the Architecture Committee election, where we have three seats available. Thus, we shall enter the Q&A and subsequently the voting phases. Candidates' names and links to their github submissions are below. The next period (September 18th - 27th) is for asking questions to candidates via this mailing list; it's an opportunity to get familiar with the perspective they would bring as an Arch Committee member before voting opens on September 28th. I'll kick it off with the first question for the candidates: What do you think is the most important feature missing from Kata Containers that will help the project increase adoption in the next year? Candidates (in alphabetical order): - Archana Shinde [1] - Eric Ernst [2] - Fabiano Fidêncio [3] - Justin He [4] [1] https://github.com/kata-containers/community/blob/master/elections/arch-comm... [2] https://github.com/kata-containers/community/blob/master/elections/arch-comm... [3] https://github.com/kata-containers/community/blob/master/elections/arch-comm... [4] https://github.com/kata-containers/community/blob/master/elections/arch-comm... Thanks! Ricardo
Community,
What do you think is the most important feature missing from Kata Containers that will help the project increase adoption in the next year?
Answering this question is not easy and I must admit a lot of back-and-forth was required from my side. There's a lot currently going on with the 2.0 about to be released and the main thing that keeps coming back to my mind is not a missing feature, but rather something that we should make considerable improvements in order to be ready for the enterprise Linux world: debuggability. Considering that others will also think the "debuggability" word is a little bit too vague, I mean: * Improve (or create) troubleshooting guides for the most commonly reported problems; - An analysis of the issues opened against the project may give a good direction of what are the pain points; * Improve (or create) tools for allowing customers of the project to "just switch a flag" and have Kata Containers running in "debug mode"; * Improve containerd-shim-kata-v2 debugging integration with the container runtime interfaces; - There's been a downgrade on the granularity of the logs with the switch to "shim v2" when compared to what we had with "shim v1". A desire for a better debuggability is a shared feeling among our community and it should be one of the points to prioritize for the coming year. Best Regards, -- Fabiano Fidêncio
Thanks Ricardo. On Fri, Sep 18, 2020 at 10:46 AM Ricardo Aravena <raravena80@gmail.com> wrote:
Hello Kata Citizens!
We have four candidate submissions for the Architecture Committee election, where we have three seats available. Thus, we shall enter the Q&A and subsequently the voting phases. Candidates' names and links to their github submissions are below.
The next period (September 18th - 27th) is for asking questions to candidates via this mailing list; it's an opportunity to get familiar with the perspective they would bring as an Arch Committee member before voting opens on September 28th.
I'll kick it off with the first question for the candidates:
What do you think is the most important feature missing from Kata Containers that will help the project increase adoption in the next year?
Similar to Fabiano's reply, I have a hard time picking one. Also similarly, I will point out things that aren't really 'features' :) I am currently looking at IO performance for storage, manageability/observability, and performance isolation (cpusets, etc). Of these, I think the manageability part is most interesting/important part that will be common among all end users. Administrators, security teams and end-users assume a certain level of visibility into the container workload that is running, which is much more challenging when leveraging the Kata Containers runtime. As a result, we need to be proactive to assure that it is easy to assess "how much is this costing?", "can I trace malicious behavior of the workload even though its in a VM?' and "how can I determine the performance profile of my application and how it is using host/guest resources." We are quite lucky that our friends at AntFinancial have done a good job pushing this forward with some of the observability APIs introduced in 2.0, but we will likely need to continue to develop this going forward, providing the right hooks for key container management tools to operate well with the Kata runtime. Of course, if our footprint and IO performance aren't kept competitive enough, then observability doesn't matter as much, so .. yeah, that's very important still too :)
Candidates (in alphabetical order): - Archana Shinde [1] - Eric Ernst [2] - Fabiano Fidêncio [3] - Justin He [4]
[1] https://github.com/kata-containers/community/blob/master/elections/arch-comm... [2] https://github.com/kata-containers/community/blob/master/elections/arch-comm... [3] https://github.com/kata-containers/community/blob/master/elections/arch-comm... [4] https://github.com/kata-containers/community/blob/master/elections/arch-comm...
Thanks! Ricardo
_______________________________________________ kata-dev mailing list kata-dev@lists.katacontainers.io http://lists.katacontainers.io/cgi-bin/mailman/listinfo/kata-dev
Hi dear candidates, It might be a bit late. May I ask another question? I have kept thinking this question for days -- What do you think about the trade-off between isolation and communication-channel between hosts and guest? Thank you folks! Cheers! Xu ------------------------------------------------------------------ From:Ricardo Aravena <raravena80@gmail.com> Send Time:2020年9月19日(星期六) 01:46 To:kata-dev@lists.katacontainers.io <kata-dev@lists.katacontainers.io> Subject:[kata-dev] Arch Committee Election: Q&A period open! Hello Kata Citizens! We have four candidate submissions for the Architecture Committee election, where we have three seats available. Thus, we shall enter the Q&A and subsequently the voting phases. Candidates' names and links to their github submissions are below. The next period (September 18th - 27th) is for asking questions to candidates via this mailing list; it's an opportunity to get familiar with the perspective they would bring as an Arch Committee member before voting opens on September 28th. I'll kick it off with the first question for the candidates: What do you think is the most important feature missing from Kata Containers that will help the project increase adoption in the next year? Candidates (in alphabetical order): - Archana Shinde [1] - Eric Ernst [2] - Fabiano Fidêncio [3] - Justin He [4] [1] https://github.com/kata-containers/community/blob/master/elections/arch-comm... [2] https://github.com/kata-containers/community/blob/master/elections/arch-comm... [3] https://github.com/kata-containers/community/blob/master/elections/arch-comm... [4] https://github.com/kata-containers/community/blob/master/elections/arch-comm... Thanks! Ricardo
Xu [...]
I have kept thinking this question for days -- What do you think about the trade-off between isolation and communication-channel between hosts and guest?
What kind of host <-> guest communication channel are we talking about? What kind of host <-> guest communication worries you? Sorry for not answering your question yet, but I'd like to understand the context better, in order to give it a proper thought before answering. Best Regards, -- Fabiano Fidêncio
Hi Fabiano,
What kind of host <-> guest communication
We could conclude our use cases to two different scenarios: - Host and guest belongs to the same tenant, who employs kata for isolation. - Host and guest belongs to different tenants, and the service provider employs kata for container (or Pod) service. For the first scenario, the users may want to setup a host <-> guest comm channel for their applications. Cheers! Xu ------------------------------------------------------------------ From:Fabiano Fidêncio <fabiano@fidencio.org> Send Time:2020年9月28日(星期一) 18:16 To:王旭(循环) <xu@hyper.sh> Cc:kata-dev@lists.katacontainers.io <kata-dev@lists.katacontainers.io> Subject:Re: [kata-dev] Arch Committee Election: Q&A period open! Xu [...]
I have kept thinking this question for days -- What do you think about the trade-off between isolation and communication-channel between hosts and guest?
What kind of host <-> guest communication channel are we talking about? What kind of host <-> guest communication worries you? Sorry for not answering your question yet, but I'd like to understand the context better, in order to give it a proper thought before answering. Best Regards, -- Fabiano Fidêncio
Thanks Ricardo, Eric and Fabiano have already talked about debuggability. I myself consider this extremely crucial for wider adaptation of Kata among our customers. I have already had discussion around this with customers using Kata to identify the gaps we have today in Kata in terms of debuggability and understand what debug features they would find beneficial while running in production scenarios. Some of this includes adding audit functionalities in Kata, making sure logs are not lost with software crashes and logs are better managed in general. I have started work around improving our debug guides and would like to improve things further by providing a Kata image with all things required for better debugging built in. That way a customer could simply switch to an image with necessary debug/tracing features baked in. The tremendous work done by Ant financial folks has really added value to Kata from observability point of view, I would like to take that forward and see how we could use it for various debug scenarios. That said, I have started looking at the shimv2 implementation for docker, and through this process I am planning to improve debuggabilty and tracing features for the shimv2 api in general. Another aspect of crucial importance is our IO performance. I have spent some efforts measuring our virtiofs performance under various stress loads. I plan to continue those efforrts in co-ordination with other Kata folks to make sure our IO performance improves in the long run. I also plan to include tests that stress our IO performance as part of a CI to see how well our IO performs. Lastly, I have worked in the past on the networking aspects of Kata to make sure Kata works with various CNI plugins and with ipv6 stacks. I wish to take this work forward to make sure Kata works well for NFV workloads by ensuring stronger support for network related features like CPU pinning and huge pages. I think this is an area where Kata could be extremely useful. Thanks, Archana From: Ricardo Aravena <raravena80@gmail.com> Date: Friday, September 18, 2020 at 10:46 AM To: kata-dev <kata-dev@lists.katacontainers.io> Subject: [kata-dev] Arch Committee Election: Q&A period open! Hello Kata Citizens! We have four candidate submissions for the Architecture Committee election, where we have three seats available. Thus, we shall enter the Q&A and subsequently the voting phases. Candidates' names and links to their github submissions are below. The next period (September 18th - 27th) is for asking questions to candidates via this mailing list; it's an opportunity to get familiar with the perspective they would bring as an Arch Committee member before voting opens on September 28th. I'll kick it off with the first question for the candidates: What do you think is the most important feature missing from Kata Containers that will help the project increase adoption in the next year? Candidates (in alphabetical order): - Archana Shinde [1] - Eric Ernst [2] - Fabiano Fidêncio [3] - Justin He [4] [1] https://github.com/kata-containers/community/blob/master/elections/arch-comm... [2] https://github.com/kata-containers/community/blob/master/elections/arch-comm... [3] https://github.com/kata-containers/community/blob/master/elections/arch-comm... [4] https://github.com/kata-containers/community/blob/master/elections/arch-comm... Thanks! Ricardo
participants (9)
-
Dr. David Alan Gilbert
-
Eric Ernst
-
Fabiano Fidêncio
-
Pradipta Kumar
-
Ricardo Aravena
-
Samuel Ortiz
-
Shinde, Archana M
-
Whaley, Graham
-
Xu Wang