Should startup and RSS performance be a working group

[Emmanuel Bernard]

With our conversations around RSS usage and build time slow creep.

And our culprits being classloading,c lass generation, configuration creep and a few more,

Should we have a slow burg WG or a WG we decide to activate in one or two releases from now to attack the problem and do a spring cleaning?

Emmanuel

[Ziphii Ltd]

We appreciate you reaching out. You can expect to hear back from us within 2 to 3 business days.

Ziphii Ltd
www.ziphii.com

This email is a service from Ziphii Ltd. Delivered by Zendesk

[MG9R2N-Y9DGY]

[Georgios Andrianakis]

+1

--
You received this message because you are subscribed to the Google Groups "Quarkus Development mailing list" group.
To unsubscribe from this group and stop receiving emails from it, send an email to quarkus-dev...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/quarkus-dev/CANYWk7NbNgT3arovAnVGr91vE8xo5oN1Zi-VJef9OLQ4X284AA%40mail.gmail.com.

[Francesco Nigro]

+1
it is kind of tied to the Holly effort on performance testing since if we enable profiling there...it will be easier

[clement escoffier]

Hello,

A working group should have a “short” life time (several months, ideally no more than 6). Here, I believe this is more of a long-running task, without a clear definition of 'done'. It’s the second time in a few weeks that I've received such a “long-running” working group request. We already do that for LTS (even if the LTS support period is defined). 

It appears more akin to a structure that would coordinate multiple short-term working groups, rather than a regular working group.  But these structures serve a different purpose. They are not there to highlight where we are going, but primarily coordinate work. 

I agree that we need to structure our effort on this topic. I would prefer to start with short, actionable working groups before we create such a structure.

Clement

To view this discussion visit https://groups.google.com/d/msgid/quarkus-dev/CALeTM-mdi4b8TK23TLrB4-Rt%3DC8mvScZBD6R%3DKN4mj7E250t3Q%40mail.gmail.com.

[Francesco Nigro]

I would start with some benchmark which reveal deficiences and create working groups to fix such i.e. benchmark-oriented - assuming a relevant benchmark (I can help to validate/verify it from the perf pov)
Or just we can list what are the aspects we know already there are deficiences e.g. classloader RSS due to linkage errors and create WG to fix such (this can be vertical to different component/extension types too) - but having a proper benchmark/microbenchmar w profiling data is key to have a validation mechanism

[Emmanuel Bernard]

I was thinking as a one off here, not a recurring long running one. Recurring as in reawaken every year or so is a possibility but I feel that it's orthogonal to the concrete conversation that got started Tuesday.

To view this discussion visit https://groups.google.com/d/msgid/quarkus-dev/C3B43A19-DF73-4A6E-AAF1-269ECE44C91C%40gmail.com.