xfs
[Top] [All Lists]

Re: [PATCH 4/4] mm: numa: Slow PTE scan rate if migration failures occur

To: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>
Subject: Re: [PATCH 4/4] mm: numa: Slow PTE scan rate if migration failures occur
From: Mel Gorman <mgorman@xxxxxxx>
Date: Fri, 20 Mar 2015 09:56:06 +0000
Cc: Dave Chinner <david@xxxxxxxxxxxxx>, Ingo Molnar <mingo@xxxxxxxxxx>, Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>, Aneesh Kumar <aneesh.kumar@xxxxxxxxxxxxxxxxxx>, Linux Kernel Mailing List <linux-kernel@xxxxxxxxxxxxxxx>, Linux-MM <linux-mm@xxxxxxxxx>, xfs@xxxxxxxxxxx, ppc-dev <linuxppc-dev@xxxxxxxxxxxxxxxx>
Delivered-to: xfs@xxxxxxxxxxx
In-reply-to: <CA+55aFy5UeNnFUTi619cs3b9Up2NQ1wbuyvcCS614+o3=z=wBQ@xxxxxxxxxxxxxx>
References: <20150317070655.GB10105@dastard> <CA+55aFzdLnFdku-gnm3mGbeS=QauYBNkFQKYXJAGkrMd2jKXhw@xxxxxxxxxxxxxx> <20150317205104.GA28621@dastard> <CA+55aFzSPcNgxw4GC7aAV1r0P5LniyVVC66COz=3cgMcx73Nag@xxxxxxxxxxxxxx> <20150317220840.GC28621@dastard> <CA+55aFwne-fe_Gg-_GTUo+iOAbbNpLBa264JqSFkH79EULyAqw@xxxxxxxxxxxxxx> <CA+55aFy-Mw74rAdLMMMUgnsG3ZttMWVNGz7CXZJY7q9fqyRYfg@xxxxxxxxxxxxxx> <CA+55aFyxA9u2cVzV+S7TSY9ZvRXCX=z22YAbi9mdPVBKmqgR5g@xxxxxxxxxxxxxx> <20150319224143.GI10105@dastard> <CA+55aFy5UeNnFUTi619cs3b9Up2NQ1wbuyvcCS614+o3=z=wBQ@xxxxxxxxxxxxxx>
User-agent: Mutt/1.5.21 (2010-09-15)
On Thu, Mar 19, 2015 at 04:05:46PM -0700, Linus Torvalds wrote:
> On Thu, Mar 19, 2015 at 3:41 PM, Dave Chinner <david@xxxxxxxxxxxxx> wrote:
> >
> > My recollection wasn't faulty - I pulled it from an earlier email.
> > That said, the original measurement might have been faulty. I ran
> > the numbers again on the 3.19 kernel I saved away from the original
> > testing. That came up at 235k, which is pretty much the same as
> > yesterday's test. The runtime,however, is unchanged from my original
> > measurements of 4m54s (pte_hack came in at 5m20s).
> 
> Ok. Good. So the "more than an order of magnitude difference" was
> really about measurement differences, not quite as real. Looks like
> more a "factor of two" than a factor of 20.
> 
> Did you do the profiles the same way? Because that would explain the
> differences in the TLB flush percentages too (the "1.4% from
> tlb_invalidate_range()" vs "pretty much everything from migration").
> 
> The runtime variation does show that there's some *big* subtle
> difference for the numa balancing in the exact TNF_NO_GROUP details.

TNF_NO_GROUP affects whether the scheduler tries to group related processes
together. Whether migration occurs depends on what node a process is
scheduled on. If processes are aggressively grouped inappropriately then it
is possible there is a bug that causes the load balancer to move processes
off a node (possible migration) with NUMA balancing trying to pull it back
(another possible migration). Small bugs there can result in excessive
migration.

-- 
Mel Gorman
SUSE Labs

<Prev in Thread] Current Thread [Next in Thread>