linux-mm.kvack.org archive mirror
 help / color / mirror / Atom feed
From: Andrea Arcangeli <aarcange@redhat.com>
To: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@techsingularity.net>,
	Andrew Morton <akpm@linux-foundation.org>,
	Linux-MM <linux-mm@kvack.org>, Rik van Riel <riel@surriel.com>,
	Johannes Weiner <hannes@cmpxchg.org>,
	Minchan Kim <minchan@kernel.org>,
	Joonsoo Kim <iamjoonsoo.kim@lge.com>,
	LKML <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH 00/34] Move LRU page reclaim from zones to nodes v9
Date: Fri, 19 Aug 2016 15:55:15 +0200	[thread overview]
Message-ID: <20160819135515.hft4t5q27za6eui2@redhat.com> (raw)
In-Reply-To: <5a560eab-cdf9-1961-1216-deff50cdf494@suse.cz>

On Fri, Aug 19, 2016 at 03:23:20PM +0200, Vlastimil Babka wrote:
> What's that? Never head of this before, but sounds scary :) I thought 
> that zone_reclaim itself was rather discouraged nowadays, not a big 
> candidate for further improvement.,,

It's some fix that I tried to push upstream but wasn't merged. I kept
maintaining it because I got customers bugreport about THP causing
regressions to node_reclaim.

Hard NUMA bindings would solve that but apparently there are apps that
prefers no memory binding to allow flexible spillover, and they only
use CPU bindings only but with a strong NUMA bias provided by
node_reclaim, by shrinking the cache (and only the cache).

In any case it was a regression caused by THP because compaction
wasn't invoked. Note zone_reclaim has a synchronous more aggressive
option that blocks for write back if needed, so invoking direct
compaction there is sure ok, if it's asked on demand.

As usual it's always a tradeoff between long live and short lived
allocation so if you reserve a system for computations and you know
your allocation are very long lived it make perfect sense to be
aggressive if you tune for it.

zone_reclaim or synchronous direct compaction are obviously bad
defaults for general purpose default settings, it doesn't mean it
should be impossible to tune a system for a certain workload to run
optimal.

> Hm I'm not so sure. Are all movable allocations highmem? For example 
> Joonsoo mentions in his ZONE_CMA patchset "blockdev file cache page 
> [...] usually has __GFP_MOVABLE but not __GFP_HIGHMEM and __GFP_USER".
> Now we also have Minchan's infrastructure for arbitrary driver 
> compaction, so those will be movable, but potentially still restricted 
> to e.g. DMA32...

One option is to forbid such corner cases... and VM_WARN_ON (not a
typo :) available in my tree) if __GFP_MOVABLE is passed on lower
classzones.

The other option would be to have a per-classzone lowpfn, highpnf scan
pointers. That has some cons but hey this whole thing is a tradeoff
isn't it?

It's about the fact we're optimizing for less frequent lowmem
allocations so we can as well provide a worse compaction for lowmem
(by reducing the MOVABLE memory restricted to lower classzones like
mentioned above), but leverage the node model to have a more powerful
that crosses all zone boundaries, when the GFP_HIGHUSER is used.

I don't see why the tradeoff is valid when it comes to the LRU but not
valid when it comes to compaction and then I've to do a blind loop of
(for-each-zone-in-the-node-in-reverse { compact_zone_order(zone) })
which works worse than before and works worse than a
zone-boundary-less compaction based on the node model.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

  reply	other threads:[~2016-08-19 13:55 UTC|newest]

Thread overview: 109+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-07-08  9:34 [PATCH 00/34] Move LRU page reclaim from zones to nodes v9 Mel Gorman
2016-07-08  9:34 ` [PATCH 01/34] mm, vmstat: add infrastructure for per-node vmstats Mel Gorman
2016-08-03 19:13   ` Reza Arbab
2016-07-08  9:34 ` [PATCH 02/34] mm, vmscan: move lru_lock to the node Mel Gorman
2016-07-12 11:06   ` Balbir Singh
2016-07-12 11:18     ` Mel Gorman
2016-07-13  5:50       ` Balbir Singh
2016-07-13  8:39         ` Vlastimil Babka
2016-07-08  9:34 ` [PATCH 03/34] mm, vmscan: move LRU lists to node Mel Gorman
2016-08-04 20:59   ` James Hogan
2016-08-05  8:41     ` Mel Gorman
2016-08-05 10:52       ` James Hogan
2016-08-05 11:55         ` Mel Gorman
2016-08-05 12:02           ` James Hogan
2016-07-08  9:34 ` [PATCH 04/34] mm, mmzone: clarify the usage of zone padding Mel Gorman
2016-07-12 13:49   ` Johannes Weiner
2016-07-08  9:34 ` [PATCH 05/34] mm, vmscan: begin reclaiming pages on a per-node basis Mel Gorman
2016-07-12 13:54   ` Johannes Weiner
2016-07-14  9:19   ` Vlastimil Babka
2016-07-08  9:34 ` [PATCH 06/34] mm, vmscan: have kswapd only scan based on the highest requested zone Mel Gorman
2016-07-12 14:05   ` Johannes Weiner
2016-07-13  8:37     ` Mel Gorman
2016-07-08  9:34 ` [PATCH 07/34] mm, vmscan: make kswapd reclaim in terms of nodes Mel Gorman
2016-08-29  9:38   ` Srikar Dronamraju
2016-08-30 12:07     ` Mel Gorman
2016-08-30 14:25       ` Srikar Dronamraju
2016-08-30 15:00         ` Mel Gorman
2016-08-31  6:09           ` Srikar Dronamraju
2016-08-31  8:49             ` Mel Gorman
2016-08-31 11:09               ` Michal Hocko
2016-08-31 12:46                 ` Mel Gorman
2016-08-31 17:33               ` Srikar Dronamraju
2016-07-08  9:34 ` [PATCH 08/34] mm, vmscan: remove balance gap Mel Gorman
2016-07-12 14:06   ` Johannes Weiner
2016-07-08  9:34 ` [PATCH 09/34] mm, vmscan: simplify the logic deciding whether kswapd sleeps Mel Gorman
2016-07-08  9:34 ` [PATCH 10/34] mm, vmscan: by default have direct reclaim only shrink once per node Mel Gorman
2016-07-08  9:34 ` [PATCH 11/34] mm, vmscan: remove duplicate logic clearing node congestion and dirty state Mel Gorman
2016-07-12 14:22   ` Johannes Weiner
2016-07-13  8:40     ` Mel Gorman
2016-07-14  9:45   ` Vlastimil Babka
2016-07-08  9:34 ` [PATCH 12/34] mm: vmscan: do not reclaim from kswapd if there is any eligible zone Mel Gorman
2016-07-12 14:29   ` Johannes Weiner
2016-07-13  8:47     ` Mel Gorman
2016-07-13 12:28       ` Johannes Weiner
2016-07-08  9:34 ` [PATCH 13/34] mm, vmscan: make shrink_node decisions more node-centric Mel Gorman
2016-07-12 14:32   ` Johannes Weiner
2016-07-13  8:48     ` Mel Gorman
2016-07-08  9:34 ` [PATCH 14/34] mm, memcg: move memcg limit enforcement from zones to nodes Mel Gorman
2016-07-12 14:38   ` Johannes Weiner
2016-07-08  9:34 ` [PATCH 15/34] mm, workingset: make working set detection node-aware Mel Gorman
2016-07-08  9:34 ` [PATCH 16/34] mm, page_alloc: consider dirtyable memory in terms of nodes Mel Gorman
2016-07-08  9:34 ` [PATCH 17/34] mm: move page mapped accounting to the node Mel Gorman
2016-07-12 14:42   ` Johannes Weiner
2016-07-08  9:34 ` [PATCH 18/34] mm: rename NR_ANON_PAGES to NR_ANON_MAPPED Mel Gorman
2016-07-12 14:58   ` Johannes Weiner
2016-07-13  8:55     ` Mel Gorman
2016-07-13 13:04       ` Johannes Weiner
2016-07-13 13:37         ` Mel Gorman
2016-07-13 21:13           ` Andrew Morton
2016-07-15 10:46             ` Mel Gorman
2016-07-15 22:35               ` Andrew Morton
2016-07-18 13:34                 ` Johannes Weiner
2016-07-14  1:27           ` Minchan Kim
2016-07-08  9:34 ` [PATCH 19/34] mm: move most file-based accounting to the node Mel Gorman
2016-07-12 15:11   ` Johannes Weiner
2016-07-08  9:34 ` [PATCH 20/34] mm: move vmscan writes and file write " Mel Gorman
2016-07-12 15:15   ` Johannes Weiner
2016-07-08  9:34 ` [PATCH 21/34] mm, vmscan: only wakeup kswapd once per node for the requested classzone Mel Gorman
2016-07-12 17:18   ` Johannes Weiner
2016-07-08  9:34 ` [PATCH 22/34] mm, page_alloc: wake kswapd based on the highest eligible zone Mel Gorman
2016-07-12 17:24   ` Johannes Weiner
2016-07-14 10:05   ` Vlastimil Babka
2016-07-08  9:34 ` [PATCH 23/34] mm: convert zone_reclaim to node_reclaim Mel Gorman
2016-07-12 17:28   ` Johannes Weiner
2016-07-08  9:35 ` [PATCH 24/34] mm, vmscan: avoid passing in classzone_idx unnecessarily to shrink_node Mel Gorman
2016-07-12 17:31   ` Johannes Weiner
2016-07-14 10:09   ` Vlastimil Babka
2016-07-08  9:35 ` [PATCH 25/34] mm, vmscan: avoid passing in classzone_idx unnecessarily to compaction_ready Mel Gorman
2016-07-12 18:01   ` Johannes Weiner
2016-07-14 12:12   ` Vlastimil Babka
2016-07-08  9:35 ` [PATCH 26/34] mm, vmscan: avoid passing in remaining unnecessarily to prepare_kswapd_sleep Mel Gorman
2016-07-12 18:06   ` Johannes Weiner
2016-07-14 12:48   ` Vlastimil Babka
2016-07-08  9:35 ` [PATCH 27/34] mm, vmscan: Have kswapd reclaim from all zones if reclaiming and buffer_heads_over_limit Mel Gorman
2016-07-12 18:10   ` Johannes Weiner
2016-07-14 12:54   ` Vlastimil Babka
2016-07-08  9:35 ` [PATCH 28/34] mm, vmscan: add classzone information to tracepoints Mel Gorman
2016-07-12 18:13   ` Johannes Weiner
2016-07-08  9:35 ` [PATCH 29/34] mm, page_alloc: remove fair zone allocation policy Mel Gorman
2016-07-12 18:18   ` Johannes Weiner
2016-07-08  9:35 ` [PATCH 30/34] mm: page_alloc: cache the last node whose dirty limit is reached Mel Gorman
2016-07-12 18:43   ` Johannes Weiner
2016-07-08  9:35 ` [PATCH 31/34] mm: vmstat: replace __count_zone_vm_events with a zone id equivalent Mel Gorman
2016-07-12 19:10   ` Johannes Weiner
2016-07-08  9:35 ` [PATCH 32/34] mm: vmstat: account per-zone stalls and pages skipped during reclaim Mel Gorman
2016-07-12 19:06   ` Johannes Weiner
2016-07-08  9:35 ` [PATCH 33/34] mm, vmstat: print node-based stats in zoneinfo file Mel Gorman
2016-07-12 19:18   ` Johannes Weiner
2016-07-14 12:56   ` Vlastimil Babka
2016-07-08  9:35 ` [PATCH 34/34] mm, vmstat: remove zone and node double accounting by approximating retries Mel Gorman
2016-07-14 13:40   ` Vlastimil Babka
2016-07-15  7:48     ` Mel Gorman
2016-07-15 12:20       ` Vlastimil Babka
2016-08-19 13:12 ` [PATCH 00/34] Move LRU page reclaim from zones to nodes v9 Andrea Arcangeli
2016-08-19 13:23   ` Vlastimil Babka
2016-08-19 13:55     ` Andrea Arcangeli [this message]
2016-08-19 14:53   ` Mel Gorman
2016-08-19 15:32     ` Andrea Arcangeli
2016-08-19 15:55       ` Mel Gorman

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20160819135515.hft4t5q27za6eui2@redhat.com \
    --to=aarcange@redhat.com \
    --cc=akpm@linux-foundation.org \
    --cc=hannes@cmpxchg.org \
    --cc=iamjoonsoo.kim@lge.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=mgorman@techsingularity.net \
    --cc=minchan@kernel.org \
    --cc=riel@surriel.com \
    --cc=vbabka@suse.cz \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body. 
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).