Re: Current ULE state and plans (was Re: HMP scheduling on FreeBSD)

Hi,

> On what basis did/do you size this cohort?

I responded to that in another mail.  This sizing is indeed very rough.  There has been only one more feedback since then (with an untold number of customers behind, and no mention of workloads, circumstances, etc.).  We'll see if we get more.

People using 4BSD, especially as they don't seem to be many of them, should feel involved in testing/reporting about ULE.

> If it's a "didin't know about it" well that becomes self-fulfilling
> if it's not talked about or otherwise advertised, doesn't it?

I understand what you mean, but it's by no means becoming self-fulfilling now, it has already been the case for more than 15 years.  I was just stating the simple fact that because 4BSD is not the default scheduler, and indeed is not advertised, there probably aren't many people using it, so it has been de facto much less tested in general in the past years than ULE.

I have been around since ULE's becoming the default scheduler, and as I understood it it was supposed to just replace 4BSD entirely, with only advantages.  It turned out it performed less well in some scenarios.  Although a lot of progress has been made, there are still known problems (see my initial mail), and thus potential scenarios where 4BSD could be better.  My aim is to bridge that gap, as I think it can be done, and only then get rid of 4BSD.

Thanks and regards.

-- 
Olivier Certner