Re: [PATCH v4] xen: sched: convert RTDS from time to event driven model

[Dario Faggioli <dario.faggioli@citrix.com>]

[-- Attachment #1.1: Type: text/plain, Size: 2185 bytes --]

On Tue, 2016-02-02 at 22:33 -0500, Meng Xu wrote:
> On Tue, Feb 2, 2016 at 10:08 AM, Dario Faggioli
> <dario.faggioli@citrix.com> wrote:
> > 
> > Is it ok to kill the replenishment in this case?
> > 
> > This is a genuine question. What does the dynamic DS algorithm that
> > you're trying to implement here says about this? Meng, maybe you
> > can
> > help here?
> 
> 
> Based on the DS algorithm (in theory), each VCPU should have its
> budget replenished at the beginning of its next period.
> 
> However, we are thinking that when a VCPU is put to sleep, no one is
> supposed to use it. Do we really need to keep updating the VCPU's
> budget? Can we just update the VCPU's budget when it is waken up
> later? This could potentially save some implementation overhead,
> IMHO.
> That's why we decided not to update the budget of VCPUs that are put
> into sleep.
> 
> > Is it ok to do this _because_ you then handle the situation of a
> > replenishment having to had happened while the vcpu was asleep in
> > rt_vcpu_wake (the '(now>=svc->cur_deadline)' thing)? Yes... it
> > probably
> > is ok for that reason...
> 
> 
> Yes, exactly. We hope this could reduce the frequency of invoking the
> replenishment timer when the system is (kind of) idle.
> 
I see what you mean. I wonder, however, given how big and complicated
this re-structuring we are doing is, whether it would not be easier to
just leave this optimization for the future, and just implement the
algorithm in the new event-driven fashion, as first step.

Early optimization is known to be a bad thing in software.

Note that I'm not saying that the optimization should happen in 6
months, it can be patch number 2 of the same series doing the event-
driven transformation... I'm just saying that we should probably have a
patch 1 which does only that, for ease of both  doing and reviewing.

Regards,
Dario
-- 
<<This happens because I choose it to happen!>> (Raistlin Majere)
-----------------------------------------------------------------
Dario Faggioli, Ph.D, http://about.me/dario.faggioli
Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)


[-- Attachment #1.2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 181 bytes --]

[-- Attachment #2: Type: text/plain, Size: 126 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel