2021-03-17 13:32:13 UTC - Keith Null: Hi, everybody~ I'm a senior student, doing my graduation project with OpenWhisk now (specifically, I need to design some new priority-based scheduling mechanism). However, I get stuck before the huge codebase of OpenWhisk and struggle to read Scala code. I just don't know where and how I should start with developing OpenWhisk. So I would really appreciate it if anyone could kindly give me some advice or reading materials. Thanks a lot in advance:pray::skin-tone-2: https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1615987933006300?thread_ts=1615987933.006300&cid=C3TPCAQG1 ---- 2021-03-17 16:39:33 UTC - Matt Rutkowski: @Dave Grove I am concerned over implications from a new issue that was opened against wskdeploy by someone I am inferring is from Apache Infra. <https://github.com/apache/openwhisk-wskdeploy/issues/1129> am further inferring that somehow the addition of “asf.yaml” (i.e., <https://github.com/apache/openwhisk-wskdeploy/pull/1128>) (plus comments the submitted made) that they are trying to build us on new servers with diff. architectures? This will impact other projects such as CLI which have never had problems in their gradle builds in these areas (i.e., gogradle plugin and use of go-bindata). Any insight, as I have no idea (nor ability) to see errors on said implied machines. https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1615999173009000?thread_ts=1615999173.009000&cid=C3TPCAQG1 ---- 2021-03-17 21:23:07 UTC - Amory Hoste: Hi all,
I was wondering if anyone could give some insights into why one would choose OpenWhisk over Knative for serverless functions from a performance / scalability perspective. I think that as of now, IBM cloud functions is still using OpenWhisk but I couldn’t find any direct comparisons between the two services. To me it would seem that using Kubernetes under the hood must have some performance implications, adding additional overheads / complexities to the scheduling and autoscaling process. Especially when scheduling hundreds or thousands of short lived functions per minute (the Kubernetes pod scheduler is centralised) or handling short bursts. I could however not find much information on this. The reason I am asking this is because I am looking into improving function latency and throughput by using different scheduling strategies as part of my master thesis but as of yet have not found compelling arguments why I would choose OpenWhisk or Knative. Any insights into why OpenWhisk / Knative might be a better choice from a scheduling perspective would also be very helpful. The advantage of OpenWhisk to me seems that all scheduling and autoscaling of functions is done in the controller rather than in two separate components - the Knative pod autoscaler and the Kubernetes pod scheduler which would mean metrics would have to be collected at two separate components to inform the scheduling decisions and would probably complicate the implementation of a new scheduler. I would very much appreciate any insights / information! https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1616016187011100?thread_ts=1616016187.011100&cid=C3TPCAQG1 ----
