From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-13.3 required=3.0 tests=BAYES_00,DKIMWL_WL_MED, DKIM_SIGNED,DKIM_VALID,DKIM_VALID_AU,HEADER_FROM_DIFFERENT_DOMAINS, MAILING_LIST_MULTI,SPF_HELO_NONE,SPF_PASS,USER_IN_DEF_DKIM_WL autolearn=no autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id 01B38C433ED for ; Wed, 14 Apr 2021 03:40:33 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by mail.kernel.org (Postfix) with ESMTP id D2F26613C0 for ; Wed, 14 Apr 2021 03:40:32 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S233280AbhDNDkv (ORCPT ); Tue, 13 Apr 2021 23:40:51 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:33052 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S233235AbhDNDkt (ORCPT ); Tue, 13 Apr 2021 23:40:49 -0400 Received: from mail-wr1-x42c.google.com (mail-wr1-x42c.google.com [IPv6:2a00:1450:4864:20::42c]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 3724EC061574 for ; Tue, 13 Apr 2021 20:40:26 -0700 (PDT) Received: by mail-wr1-x42c.google.com with SMTP id e7so9443906wrs.11 for ; Tue, 13 Apr 2021 20:40:26 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20161025; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc:content-transfer-encoding; bh=k/2wqUGNMdxVCTzDk5Rp1/XZDjJfyD4NKS6hTd5+YsA=; b=ZBTi3z/mTfpL41Y/+Y2xI9/M/wR9uG08zUJrx0LWJBlqobdryKyQ2HiJIo1CpxoDcl PoLm3xzFyYfQAlQJWl9GPZZeMupCpTwRNiGxcVVEveVX27VyrFV3I7/DwbyrNA83pdPp k0P4fTfB1Mtq+tt1OJkbGJDNUneI33Z0JWL0DvYvxA+AjjSGTZOlGig99NvOm70W51UC jtpwRwctoJLOzt103yubVNVAkmSjQhmvd2t3ipbRK5MWKJDqRAZrZ0fywEaW4rH2r4zq KZt2P/RDnAq2Yb+FLSo2yQfhqu8VG8aRhrLtJxcIWt1Ro2r2nExvGfj7d86fBWN5NiPU U0lQ== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc:content-transfer-encoding; bh=k/2wqUGNMdxVCTzDk5Rp1/XZDjJfyD4NKS6hTd5+YsA=; b=mF12StqlC0lfs3bp7IAklu6JIcHzTY+NJDxDyBJZTvL2n5WhGmJ8hHZG0t55t1S5uw 6luxTV1givdXTay794pHm0fRMTHgodVyIjWQmcmYPYIQ69GEi7nErbn3ob6imHjpCoW/ ux1acq47iHJA+e60+5XQvBY+cpMKyo+rrTReV0mQqtrpf4z7IjvLuvaa25UJQjdEQTxp nZFbJZ0l/C4dvUVyHgEJNuwrMfJEwt0ycqQ8/jFOa/GmsRVdO6QFqK3yNAAAE/Jy3TQl F0//lJY88gnEbHPBjTxDCE47IeRX8SQOC53kyFyuZWkygZaM6XfXDmUSKaj9TpkaerAW ajJg== X-Gm-Message-State: AOAM532g9pfkqXvBVivX0CQR8Ll2fc+MeEPwJhNpaVeYmPtEiyBpfPtc GEtA3XtSU57Fmqf61VEN5WX0xFpOEeGUffgR5+0SuQ== X-Google-Smtp-Source: ABdhPJzzr+xJDo7jOQU4cqfs6V3QmDUF5MHh323dLBAaoy3dXmRX2HEI8dxVX/EkwULOTyUwl5C4kSi3ThOx3daPCYI= X-Received: by 2002:a5d:6852:: with SMTP id o18mr19553478wrw.426.1618371624435; Tue, 13 Apr 2021 20:40:24 -0700 (PDT) MIME-Version: 1.0 References: <20210413075155.32652-1-sjpark@amazon.de> <3ddd4f8a-8e51-662b-df11-a63a0e75b2bc@kernel.dk> <20210413231436.GF63242@dread.disaster.area> In-Reply-To: <20210413231436.GF63242@dread.disaster.area> From: Yu Zhao Date: Tue, 13 Apr 2021 21:40:12 -0600 Message-ID: Subject: Re: [PATCH v2 00/16] Multigenerational LRU Framework To: Dave Chinner Cc: Jens Axboe , SeongJae Park , Linux-MM , Andi Kleen , Andrew Morton , Benjamin Manes , Dave Hansen , Hillf Danton , Johannes Weiner , Jonathan Corbet , Joonsoo Kim , Matthew Wilcox , Mel Gorman , Miaohe Lin , Michael Larabel , Michal Hocko , Michel Lespinasse , Rik van Riel , Roman Gushchin , Rong Chen , SeongJae Park , Tim Chen , Vlastimil Babka , Yang Shi , Ying Huang , Zi Yan , linux-kernel , lkp@lists.01.org, Kernel Page Reclaim v2 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Tue, Apr 13, 2021 at 5:14 PM Dave Chinner wrote: > > On Tue, Apr 13, 2021 at 10:13:24AM -0600, Jens Axboe wrote: > > On 4/13/21 1:51 AM, SeongJae Park wrote: > > > From: SeongJae Park > > > > > > Hello, > > > > > > > > > Very interesting work, thank you for sharing this :) > > > > > > On Tue, 13 Apr 2021 00:56:17 -0600 Yu Zhao wrote: > > > > > >> What's new in v2 > > >> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D > > >> Special thanks to Jens Axboe for reporting a regression in buffered > > >> I/O and helping test the fix. > > > > > > Is the discussion open? If so, could you please give me a link? > > > > I wasn't on the initial post (or any of the lists it was posted to), bu= t > > it's on the google page reclaim list. Not sure if that is public or not= . > > > > tldr is that I was pretty excited about this work, as buffered IO tends > > to suck (a lot) for high throughput applications. My test case was > > pretty simple: > > > > Randomly read a fast device, using 4k buffered IO, and watch what > > happens when the page cache gets filled up. For this particular test, > > we'll initially be doing 2.1GB/sec of IO, and then drop to 1.5-1.6GB/se= c > > with kswapd using a lot of CPU trying to keep up. That's mainline > > behavior. > > I see this exact same behaviour here, too, but I RCA'd it to > contention between the inode and memory reclaim for the mapping > structure that indexes the page cache. Basically the mapping tree > lock is the contention point here - you can either be adding pages > to the mapping during IO, or memory reclaim can be removing pages > from the mapping, but we can't do both at once. > > So we end up with kswapd spinning on the mapping tree lock like so > when doing 1.6GB/s in 4kB buffered IO: > > - 20.06% 0.00% [kernel] [k] kswapd = = =E2=96=92 > - 20.06% kswapd = = =E2=96=92 > - 20.05% balance_pgdat = = =E2=96=92 > - 20.03% shrink_node = = =E2=96=92 > - 19.92% shrink_lruvec = = =E2=96=92 > - 19.91% shrink_inactive_list = = =E2=96=92 > - 19.22% shrink_page_list = = =E2=96=92 > - 17.51% __remove_mapping = = =E2=96=92 > - 14.16% _raw_spin_lock_irqsave = = =E2=96=92 > - 14.14% do_raw_spin_lock = = =E2=96=92 > __pv_queued_spin_lock_slowpath = = =E2=96=92 > - 1.56% __delete_from_page_cache = = =E2=96=92 > 0.63% xas_store = = =E2=96=92 > - 0.78% _raw_spin_unlock_irqrestore = = =E2=96=92 > - 0.69% do_raw_spin_unlock = = =E2=96=92 > __raw_callee_save___pv_queued_spin_unlock= = =E2=96=92 > - 0.82% free_unref_page_list = = =E2=96=92 > - 0.72% free_unref_page_commit = = =E2=96=92 > 0.57% free_pcppages_bulk = = =E2=96=92 > > And these are the processes consuming CPU: > > 5171 root 20 0 1442496 5696 1284 R 99.7 0.0 1:07.78 fi= o > 1150 root 20 0 0 0 0 S 47.4 0.0 0:22.70 ks= wapd1 > 1146 root 20 0 0 0 0 S 44.0 0.0 0:21.85 ks= wapd0 > 1152 root 20 0 0 0 0 S 39.7 0.0 0:18.28 ks= wapd3 > 1151 root 20 0 0 0 0 S 15.2 0.0 0:12.14 ks= wapd2 > > i.e. when memory reclaim kicks in, the read process has 20% less > time with exclusive access to the mapping tree to insert new pages. > Hence buffered read performance goes down quite substantially when > memory reclaim kicks in, and this really has nothing to do with the > memory reclaim LRU scanning algorithm. > > I can actually get this machine to pin those 5 processes to 100% CPU > under certain conditions. Each process is spinning all that extra > time on the mapping tree lock, and performance degrades further. > Changing the LRU reclaim algorithm won't fix this - the workload is > solidly bound by the exclusive nature of the mapping tree lock and > the number of tasks trying to obtain it exclusively... > > > The initial posting of this patchset did no better, in fact it did a bi= t > > worse. Performance dropped to the same levels and kswapd was using as > > much CPU as before, but on top of that we also got excessive swapping. > > Not at a high rate, but 5-10MB/sec continually. > > > > I had some back and forths with Yu Zhao and tested a few new revisions, > > and the current series does much better in this regard. Performance > > still dips a bit when page cache fills, but not nearly as much, and > > kswapd is using less CPU than before. > > Profiles would be interesting, because it sounds to me like reclaim > *might* be batching page cache removal better (e.g. fewer, larger > batches) and so spending less time contending on the mapping tree > lock... > > IOWs, I suspect this result might actually be a result of less lock > contention due to a change in batch processing characteristics of > the new algorithm rather than it being a "better" algorithm... I appreciate the profile. But there is no batching in __remove_mapping() -- it locks the mapping for each page, and therefore the lock contention penalizes the mainline and this patchset equally. It looks worse on your system because the four kswapd threads from different nodes were working on the same file. And kswapd is only one of two paths that could affect the performance. The kernel context of the test process is where the improvement mainly comes from. I also suspect you were testing a file much larger than your memory size. If so, sorry to tell you that a file only a few times larger, e.g. twice, would be worse. Here is my take: Claim ----- This patchset is a "better" algorithm. (Technically it's not an algorithm, it's a feedback loop.) Theoretical basis ----------------- An open-loop control (the mainline) can only be better if the margin of error in its prediction of the future events is less than that from the trial-and-error of a closed-loop control (this patchset). For simple machines, it surely can. For page reclaim, AFAIK, it can't. A typical example: when randomly accessing a (not infinitely) large file via buffered io long enough, we're bound to hit the same blocks multiple times. Should we activate the pages containing those blocks, i.e., to move them to the active lru list? No. RCA --- For the fio/io_uring benchmark, the "No" is the key. The mainline activates pages accessed multiple times. This is done in the buffered io access path by mark_page_accessed(), and it takes the lru lock, which is contended under memory pressure. This contention slows down both the access path and kswapd. But kswapd is not the problem here because we are measuring the io_uring process, not kswap. For this patchset, there are no activations since the refault rates of pages accessed multiple times are similar to those accessed only once -- activations will only be done to pages from tiers with higher refault rates. If you wish to debunk --------------------- git fetch https://linux-mm.googlesource.com/page-reclaim refs/changes/73/11= 73/1 CONFIG_LRU_GEN=3Dy CONFIG_LRU_GEN_ENABLED=3Dy Run your benchmarks Profiles (200G mem + 400G file) ------------------------------- A quick test from Jens' fio/io_uring: -rc7 13.30% io_uring xas_load 13.22% io_uring _copy_to_iter 12.30% io_uring __add_to_page_cache_locked 7.43% io_uring clear_page_erms 4.18% io_uring filemap_get_read_batch 3.54% io_uring get_page_from_freelist 2.98% io_uring ***native_queued_spin_lock_slowpath*** 1.61% io_uring page_cache_ra_unbounded 1.16% io_uring xas_start 1.08% io_uring filemap_read 1.07% io_uring ***__activate_page*** lru lock: 2.98% (lru addition + activation) activation: 1.07% -rc7 + this patchset 14.44% io_uring xas_load 14.14% io_uring _copy_to_iter 11.15% io_uring __add_to_page_cache_locked 6.56% io_uring clear_page_erms 4.44% io_uring filemap_get_read_batch 2.14% io_uring get_page_from_freelist 1.32% io_uring page_cache_ra_unbounded 1.20% io_uring psi_group_change 1.18% io_uring filemap_read 1.09% io_uring ****native_queued_spin_lock_slowpath**** 1.08% io_uring do_mpage_readpage lru lock: 1.09% (lru addition only) And I plan to reach out to other communities, e.g., PostgreSQL, to benchmark the patchset. I heard they have been complaining about the buffered io performance under memory pressure. Any other benchmarks you'd suggest? BTW, you might find another surprise in how less frequently slab shrinkers are called under memory pressure, because this patchset is a lot better at finding pages to reclaim and therefore doesn't overkill slabs. Thanks. From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-13.3 required=3.0 tests=BAYES_00,DKIMWL_WL_MED, DKIM_SIGNED,DKIM_VALID,DKIM_VALID_AU,HEADER_FROM_DIFFERENT_DOMAINS, MAILING_LIST_MULTI,SPF_HELO_NONE,SPF_PASS,USER_IN_DEF_DKIM_WL autolearn=no autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id 37AE5C433ED for ; Wed, 14 Apr 2021 03:40:28 +0000 (UTC) Received: from kanga.kvack.org (kanga.kvack.org [205.233.56.17]) by mail.kernel.org (Postfix) with ESMTP id 94302613BB for ; Wed, 14 Apr 2021 03:40:27 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org 94302613BB Authentication-Results: mail.kernel.org; dmarc=fail (p=reject dis=none) header.from=google.com Authentication-Results: mail.kernel.org; spf=pass smtp.mailfrom=owner-linux-mm@kvack.org Received: by kanga.kvack.org (Postfix) id EEC846B0072; Tue, 13 Apr 2021 23:40:26 -0400 (EDT) Received: by kanga.kvack.org (Postfix, from userid 40) id E9D756B0073; Tue, 13 Apr 2021 23:40:26 -0400 (EDT) X-Delivered-To: int-list-linux-mm@kvack.org Received: by kanga.kvack.org (Postfix, from userid 63042) id D16476B0074; Tue, 13 Apr 2021 23:40:26 -0400 (EDT) X-Delivered-To: linux-mm@kvack.org Received: from forelay.hostedemail.com (smtprelay0136.hostedemail.com [216.40.44.136]) by kanga.kvack.org (Postfix) with ESMTP id B3ABD6B0072 for ; Tue, 13 Apr 2021 23:40:26 -0400 (EDT) Received: from smtpin27.hostedemail.com (10.5.19.251.rfc1918.com [10.5.19.251]) by forelay02.hostedemail.com (Postfix) with ESMTP id 62055448B9C3 for ; Wed, 14 Apr 2021 03:40:26 +0000 (UTC) X-FDA: 78029570052.27.2343E54 Received: from mail-wr1-f49.google.com (mail-wr1-f49.google.com [209.85.221.49]) by imf07.hostedemail.com (Postfix) with ESMTP id A4374A000390 for ; Wed, 14 Apr 2021 03:40:25 +0000 (UTC) Received: by mail-wr1-f49.google.com with SMTP id w4so14675773wrt.5 for ; Tue, 13 Apr 2021 20:40:25 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20161025; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc:content-transfer-encoding; bh=k/2wqUGNMdxVCTzDk5Rp1/XZDjJfyD4NKS6hTd5+YsA=; b=ZBTi3z/mTfpL41Y/+Y2xI9/M/wR9uG08zUJrx0LWJBlqobdryKyQ2HiJIo1CpxoDcl PoLm3xzFyYfQAlQJWl9GPZZeMupCpTwRNiGxcVVEveVX27VyrFV3I7/DwbyrNA83pdPp k0P4fTfB1Mtq+tt1OJkbGJDNUneI33Z0JWL0DvYvxA+AjjSGTZOlGig99NvOm70W51UC jtpwRwctoJLOzt103yubVNVAkmSjQhmvd2t3ipbRK5MWKJDqRAZrZ0fywEaW4rH2r4zq KZt2P/RDnAq2Yb+FLSo2yQfhqu8VG8aRhrLtJxcIWt1Ro2r2nExvGfj7d86fBWN5NiPU U0lQ== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc:content-transfer-encoding; bh=k/2wqUGNMdxVCTzDk5Rp1/XZDjJfyD4NKS6hTd5+YsA=; b=tEAt9vdy7KXNrKkFXgEKBGyMxb/sb9TXemQmBDXG5Xd/kYDnKLuhk1uhwMVvEzaR9e 4ufCjUAYXsrowwcQ1ZYJePfiW6+mPbca1ErL7NX2g5Lg99VE+iZJqkSopGOgfjI13FGt WvWwJQRM61WPzRovV6inF+ytSnT11QClf0bngpXZUV4g9mRTASTxu/VTCPyIFldzsQlK jjNG9JwBUvKnRQnbwKpitmLnfOZha2/g8nyHH/J3DRKkCRnruU9TXmV11TxLvqeEsYm7 t60Bw22Os9GWP7Is2xbRa+txurDpNPQYZLws8iTEq5/VZHMa/HZ697Xt/T4FlQeMJ39N +VGg== X-Gm-Message-State: AOAM531GOwMwkaGARlFTtpqpcSUnVPENj+Nygg5BuaeBfHUMo7PfrZvN 9TJaC549kTFOehluq5MpNAgPtLf/pYPWH6B+X8/A0w== X-Google-Smtp-Source: ABdhPJzzr+xJDo7jOQU4cqfs6V3QmDUF5MHh323dLBAaoy3dXmRX2HEI8dxVX/EkwULOTyUwl5C4kSi3ThOx3daPCYI= X-Received: by 2002:a5d:6852:: with SMTP id o18mr19553478wrw.426.1618371624435; Tue, 13 Apr 2021 20:40:24 -0700 (PDT) MIME-Version: 1.0 References: <20210413075155.32652-1-sjpark@amazon.de> <3ddd4f8a-8e51-662b-df11-a63a0e75b2bc@kernel.dk> <20210413231436.GF63242@dread.disaster.area> In-Reply-To: <20210413231436.GF63242@dread.disaster.area> From: Yu Zhao Date: Tue, 13 Apr 2021 21:40:12 -0600 Message-ID: Subject: Re: [PATCH v2 00/16] Multigenerational LRU Framework To: Dave Chinner Cc: Jens Axboe , SeongJae Park , Linux-MM , Andi Kleen , Andrew Morton , Benjamin Manes , Dave Hansen , Hillf Danton , Johannes Weiner , Jonathan Corbet , Joonsoo Kim , Matthew Wilcox , Mel Gorman , Miaohe Lin , Michael Larabel , Michal Hocko , Michel Lespinasse , Rik van Riel , Roman Gushchin , Rong Chen , SeongJae Park , Tim Chen , Vlastimil Babka , Yang Shi , Ying Huang , Zi Yan , linux-kernel , lkp@lists.01.org, Kernel Page Reclaim v2 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable X-Rspamd-Queue-Id: A4374A000390 X-Stat-Signature: 6t5etm7k3zy4dbpdirsqoe8m1rt9btid X-Rspamd-Server: rspam02 Received-SPF: none (google.com>: No applicable sender policy available) receiver=imf07; identity=mailfrom; envelope-from=""; helo=mail-wr1-f49.google.com; client-ip=209.85.221.49 X-HE-DKIM-Result: pass/pass X-HE-Tag: 1618371625-276085 X-Bogosity: Ham, tests=bogofilter, spamicity=0.000000, version=1.2.4 Sender: owner-linux-mm@kvack.org Precedence: bulk X-Loop: owner-majordomo@kvack.org List-ID: On Tue, Apr 13, 2021 at 5:14 PM Dave Chinner wrote: > > On Tue, Apr 13, 2021 at 10:13:24AM -0600, Jens Axboe wrote: > > On 4/13/21 1:51 AM, SeongJae Park wrote: > > > From: SeongJae Park > > > > > > Hello, > > > > > > > > > Very interesting work, thank you for sharing this :) > > > > > > On Tue, 13 Apr 2021 00:56:17 -0600 Yu Zhao wrote: > > > > > >> What's new in v2 > > >> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D > > >> Special thanks to Jens Axboe for reporting a regression in buffered > > >> I/O and helping test the fix. > > > > > > Is the discussion open? If so, could you please give me a link? > > > > I wasn't on the initial post (or any of the lists it was posted to), bu= t > > it's on the google page reclaim list. Not sure if that is public or not= . > > > > tldr is that I was pretty excited about this work, as buffered IO tends > > to suck (a lot) for high throughput applications. My test case was > > pretty simple: > > > > Randomly read a fast device, using 4k buffered IO, and watch what > > happens when the page cache gets filled up. For this particular test, > > we'll initially be doing 2.1GB/sec of IO, and then drop to 1.5-1.6GB/se= c > > with kswapd using a lot of CPU trying to keep up. That's mainline > > behavior. > > I see this exact same behaviour here, too, but I RCA'd it to > contention between the inode and memory reclaim for the mapping > structure that indexes the page cache. Basically the mapping tree > lock is the contention point here - you can either be adding pages > to the mapping during IO, or memory reclaim can be removing pages > from the mapping, but we can't do both at once. > > So we end up with kswapd spinning on the mapping tree lock like so > when doing 1.6GB/s in 4kB buffered IO: > > - 20.06% 0.00% [kernel] [k] kswapd = = =E2=96=92 > - 20.06% kswapd = = =E2=96=92 > - 20.05% balance_pgdat = = =E2=96=92 > - 20.03% shrink_node = = =E2=96=92 > - 19.92% shrink_lruvec = = =E2=96=92 > - 19.91% shrink_inactive_list = = =E2=96=92 > - 19.22% shrink_page_list = = =E2=96=92 > - 17.51% __remove_mapping = = =E2=96=92 > - 14.16% _raw_spin_lock_irqsave = = =E2=96=92 > - 14.14% do_raw_spin_lock = = =E2=96=92 > __pv_queued_spin_lock_slowpath = = =E2=96=92 > - 1.56% __delete_from_page_cache = = =E2=96=92 > 0.63% xas_store = = =E2=96=92 > - 0.78% _raw_spin_unlock_irqrestore = = =E2=96=92 > - 0.69% do_raw_spin_unlock = = =E2=96=92 > __raw_callee_save___pv_queued_spin_unlock= = =E2=96=92 > - 0.82% free_unref_page_list = = =E2=96=92 > - 0.72% free_unref_page_commit = = =E2=96=92 > 0.57% free_pcppages_bulk = = =E2=96=92 > > And these are the processes consuming CPU: > > 5171 root 20 0 1442496 5696 1284 R 99.7 0.0 1:07.78 fi= o > 1150 root 20 0 0 0 0 S 47.4 0.0 0:22.70 ks= wapd1 > 1146 root 20 0 0 0 0 S 44.0 0.0 0:21.85 ks= wapd0 > 1152 root 20 0 0 0 0 S 39.7 0.0 0:18.28 ks= wapd3 > 1151 root 20 0 0 0 0 S 15.2 0.0 0:12.14 ks= wapd2 > > i.e. when memory reclaim kicks in, the read process has 20% less > time with exclusive access to the mapping tree to insert new pages. > Hence buffered read performance goes down quite substantially when > memory reclaim kicks in, and this really has nothing to do with the > memory reclaim LRU scanning algorithm. > > I can actually get this machine to pin those 5 processes to 100% CPU > under certain conditions. Each process is spinning all that extra > time on the mapping tree lock, and performance degrades further. > Changing the LRU reclaim algorithm won't fix this - the workload is > solidly bound by the exclusive nature of the mapping tree lock and > the number of tasks trying to obtain it exclusively... > > > The initial posting of this patchset did no better, in fact it did a bi= t > > worse. Performance dropped to the same levels and kswapd was using as > > much CPU as before, but on top of that we also got excessive swapping. > > Not at a high rate, but 5-10MB/sec continually. > > > > I had some back and forths with Yu Zhao and tested a few new revisions, > > and the current series does much better in this regard. Performance > > still dips a bit when page cache fills, but not nearly as much, and > > kswapd is using less CPU than before. > > Profiles would be interesting, because it sounds to me like reclaim > *might* be batching page cache removal better (e.g. fewer, larger > batches) and so spending less time contending on the mapping tree > lock... > > IOWs, I suspect this result might actually be a result of less lock > contention due to a change in batch processing characteristics of > the new algorithm rather than it being a "better" algorithm... I appreciate the profile. But there is no batching in __remove_mapping() -- it locks the mapping for each page, and therefore the lock contention penalizes the mainline and this patchset equally. It looks worse on your system because the four kswapd threads from different nodes were working on the same file. And kswapd is only one of two paths that could affect the performance. The kernel context of the test process is where the improvement mainly comes from. I also suspect you were testing a file much larger than your memory size. If so, sorry to tell you that a file only a few times larger, e.g. twice, would be worse. Here is my take: Claim ----- This patchset is a "better" algorithm. (Technically it's not an algorithm, it's a feedback loop.) Theoretical basis ----------------- An open-loop control (the mainline) can only be better if the margin of error in its prediction of the future events is less than that from the trial-and-error of a closed-loop control (this patchset). For simple machines, it surely can. For page reclaim, AFAIK, it can't. A typical example: when randomly accessing a (not infinitely) large file via buffered io long enough, we're bound to hit the same blocks multiple times. Should we activate the pages containing those blocks, i.e., to move them to the active lru list? No. RCA --- For the fio/io_uring benchmark, the "No" is the key. The mainline activates pages accessed multiple times. This is done in the buffered io access path by mark_page_accessed(), and it takes the lru lock, which is contended under memory pressure. This contention slows down both the access path and kswapd. But kswapd is not the problem here because we are measuring the io_uring process, not kswap. For this patchset, there are no activations since the refault rates of pages accessed multiple times are similar to those accessed only once -- activations will only be done to pages from tiers with higher refault rates. If you wish to debunk --------------------- git fetch https://linux-mm.googlesource.com/page-reclaim refs/changes/73/11= 73/1 CONFIG_LRU_GEN=3Dy CONFIG_LRU_GEN_ENABLED=3Dy Run your benchmarks Profiles (200G mem + 400G file) ------------------------------- A quick test from Jens' fio/io_uring: -rc7 13.30% io_uring xas_load 13.22% io_uring _copy_to_iter 12.30% io_uring __add_to_page_cache_locked 7.43% io_uring clear_page_erms 4.18% io_uring filemap_get_read_batch 3.54% io_uring get_page_from_freelist 2.98% io_uring ***native_queued_spin_lock_slowpath*** 1.61% io_uring page_cache_ra_unbounded 1.16% io_uring xas_start 1.08% io_uring filemap_read 1.07% io_uring ***__activate_page*** lru lock: 2.98% (lru addition + activation) activation: 1.07% -rc7 + this patchset 14.44% io_uring xas_load 14.14% io_uring _copy_to_iter 11.15% io_uring __add_to_page_cache_locked 6.56% io_uring clear_page_erms 4.44% io_uring filemap_get_read_batch 2.14% io_uring get_page_from_freelist 1.32% io_uring page_cache_ra_unbounded 1.20% io_uring psi_group_change 1.18% io_uring filemap_read 1.09% io_uring ****native_queued_spin_lock_slowpath**** 1.08% io_uring do_mpage_readpage lru lock: 1.09% (lru addition only) And I plan to reach out to other communities, e.g., PostgreSQL, to benchmark the patchset. I heard they have been complaining about the buffered io performance under memory pressure. Any other benchmarks you'd suggest? BTW, you might find another surprise in how less frequently slab shrinkers are called under memory pressure, because this patchset is a lot better at finding pages to reclaim and therefore doesn't overkill slabs. Thanks. From mboxrd@z Thu Jan 1 00:00:00 1970 Content-Type: multipart/mixed; boundary="===============0048406909259738833==" MIME-Version: 1.0 From: Yu Zhao To: lkp@lists.01.org Subject: Re: [PATCH v2 00/16] Multigenerational LRU Framework Date: Tue, 13 Apr 2021 21:40:12 -0600 Message-ID: In-Reply-To: <20210413231436.GF63242@dread.disaster.area> List-Id: --===============0048406909259738833== Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable On Tue, Apr 13, 2021 at 5:14 PM Dave Chinner wrote: > > On Tue, Apr 13, 2021 at 10:13:24AM -0600, Jens Axboe wrote: > > On 4/13/21 1:51 AM, SeongJae Park wrote: > > > From: SeongJae Park > > > > > > Hello, > > > > > > > > > Very interesting work, thank you for sharing this :) > > > > > > On Tue, 13 Apr 2021 00:56:17 -0600 Yu Zhao wrot= e: > > > > > >> What's new in v2 > > >> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D > > >> Special thanks to Jens Axboe for reporting a regression in buffered > > >> I/O and helping test the fix. > > > > > > Is the discussion open? If so, could you please give me a link? > > > > I wasn't on the initial post (or any of the lists it was posted to), but > > it's on the google page reclaim list. Not sure if that is public or not. > > > > tldr is that I was pretty excited about this work, as buffered IO tends > > to suck (a lot) for high throughput applications. My test case was > > pretty simple: > > > > Randomly read a fast device, using 4k buffered IO, and watch what > > happens when the page cache gets filled up. For this particular test, > > we'll initially be doing 2.1GB/sec of IO, and then drop to 1.5-1.6GB/sec > > with kswapd using a lot of CPU trying to keep up. That's mainline > > behavior. > > I see this exact same behaviour here, too, but I RCA'd it to > contention between the inode and memory reclaim for the mapping > structure that indexes the page cache. Basically the mapping tree > lock is the contention point here - you can either be adding pages > to the mapping during IO, or memory reclaim can be removing pages > from the mapping, but we can't do both at once. > > So we end up with kswapd spinning on the mapping tree lock like so > when doing 1.6GB/s in 4kB buffered IO: > > - 20.06% 0.00% [kernel] [k] kswapd = = =E2=96=92 > - 20.06% kswapd = = =E2=96=92 > - 20.05% balance_pgdat = = =E2=96=92 > - 20.03% shrink_node = = =E2=96=92 > - 19.92% shrink_lruvec = = =E2=96=92 > - 19.91% shrink_inactive_list = = =E2=96=92 > - 19.22% shrink_page_list = = =E2=96=92 > - 17.51% __remove_mapping = = =E2=96=92 > - 14.16% _raw_spin_lock_irqsave = = =E2=96=92 > - 14.14% do_raw_spin_lock = = =E2=96=92 > __pv_queued_spin_lock_slowpath = = =E2=96=92 > - 1.56% __delete_from_page_cache = = =E2=96=92 > 0.63% xas_store = = =E2=96=92 > - 0.78% _raw_spin_unlock_irqrestore = = =E2=96=92 > - 0.69% do_raw_spin_unlock = = =E2=96=92 > __raw_callee_save___pv_queued_spin_unlock= = =E2=96=92 > - 0.82% free_unref_page_list = = =E2=96=92 > - 0.72% free_unref_page_commit = = =E2=96=92 > 0.57% free_pcppages_bulk = = =E2=96=92 > > And these are the processes consuming CPU: > > 5171 root 20 0 1442496 5696 1284 R 99.7 0.0 1:07.78 fio > 1150 root 20 0 0 0 0 S 47.4 0.0 0:22.70 ks= wapd1 > 1146 root 20 0 0 0 0 S 44.0 0.0 0:21.85 ks= wapd0 > 1152 root 20 0 0 0 0 S 39.7 0.0 0:18.28 ks= wapd3 > 1151 root 20 0 0 0 0 S 15.2 0.0 0:12.14 ks= wapd2 > > i.e. when memory reclaim kicks in, the read process has 20% less > time with exclusive access to the mapping tree to insert new pages. > Hence buffered read performance goes down quite substantially when > memory reclaim kicks in, and this really has nothing to do with the > memory reclaim LRU scanning algorithm. > > I can actually get this machine to pin those 5 processes to 100% CPU > under certain conditions. Each process is spinning all that extra > time on the mapping tree lock, and performance degrades further. > Changing the LRU reclaim algorithm won't fix this - the workload is > solidly bound by the exclusive nature of the mapping tree lock and > the number of tasks trying to obtain it exclusively... > > > The initial posting of this patchset did no better, in fact it did a bit > > worse. Performance dropped to the same levels and kswapd was using as > > much CPU as before, but on top of that we also got excessive swapping. > > Not at a high rate, but 5-10MB/sec continually. > > > > I had some back and forths with Yu Zhao and tested a few new revisions, > > and the current series does much better in this regard. Performance > > still dips a bit when page cache fills, but not nearly as much, and > > kswapd is using less CPU than before. > > Profiles would be interesting, because it sounds to me like reclaim > *might* be batching page cache removal better (e.g. fewer, larger > batches) and so spending less time contending on the mapping tree > lock... > > IOWs, I suspect this result might actually be a result of less lock > contention due to a change in batch processing characteristics of > the new algorithm rather than it being a "better" algorithm... I appreciate the profile. But there is no batching in __remove_mapping() -- it locks the mapping for each page, and therefore the lock contention penalizes the mainline and this patchset equally. It looks worse on your system because the four kswapd threads from different nodes were working on the same file. And kswapd is only one of two paths that could affect the performance. The kernel context of the test process is where the improvement mainly comes from. I also suspect you were testing a file much larger than your memory size. If so, sorry to tell you that a file only a few times larger, e.g. twice, would be worse. Here is my take: Claim ----- This patchset is a "better" algorithm. (Technically it's not an algorithm, it's a feedback loop.) Theoretical basis ----------------- An open-loop control (the mainline) can only be better if the margin of error in its prediction of the future events is less than that from the trial-and-error of a closed-loop control (this patchset). For simple machines, it surely can. For page reclaim, AFAIK, it can't. A typical example: when randomly accessing a (not infinitely) large file via buffered io long enough, we're bound to hit the same blocks multiple times. Should we activate the pages containing those blocks, i.e., to move them to the active lru list? No. RCA --- For the fio/io_uring benchmark, the "No" is the key. The mainline activates pages accessed multiple times. This is done in the buffered io access path by mark_page_accessed(), and it takes the lru lock, which is contended under memory pressure. This contention slows down both the access path and kswapd. But kswapd is not the problem here because we are measuring the io_uring process, not kswap. For this patchset, there are no activations since the refault rates of pages accessed multiple times are similar to those accessed only once -- activations will only be done to pages from tiers with higher refault rates. If you wish to debunk --------------------- git fetch https://linux-mm.googlesource.com/page-reclaim refs/changes/73/11= 73/1 CONFIG_LRU_GEN=3Dy CONFIG_LRU_GEN_ENABLED=3Dy Run your benchmarks Profiles (200G mem + 400G file) ------------------------------- A quick test from Jens' fio/io_uring: -rc7 13.30% io_uring xas_load 13.22% io_uring _copy_to_iter 12.30% io_uring __add_to_page_cache_locked 7.43% io_uring clear_page_erms 4.18% io_uring filemap_get_read_batch 3.54% io_uring get_page_from_freelist 2.98% io_uring ***native_queued_spin_lock_slowpath*** 1.61% io_uring page_cache_ra_unbounded 1.16% io_uring xas_start 1.08% io_uring filemap_read 1.07% io_uring ***__activate_page*** lru lock: 2.98% (lru addition + activation) activation: 1.07% -rc7 + this patchset 14.44% io_uring xas_load 14.14% io_uring _copy_to_iter 11.15% io_uring __add_to_page_cache_locked 6.56% io_uring clear_page_erms 4.44% io_uring filemap_get_read_batch 2.14% io_uring get_page_from_freelist 1.32% io_uring page_cache_ra_unbounded 1.20% io_uring psi_group_change 1.18% io_uring filemap_read 1.09% io_uring ****native_queued_spin_lock_slowpath**** 1.08% io_uring do_mpage_readpage lru lock: 1.09% (lru addition only) And I plan to reach out to other communities, e.g., PostgreSQL, to benchmark the patchset. I heard they have been complaining about the buffered io performance under memory pressure. Any other benchmarks you'd suggest? BTW, you might find another surprise in how less frequently slab shrinkers are called under memory pressure, because this patchset is a lot better at finding pages to reclaim and therefore doesn't overkill slabs. Thanks. --===============0048406909259738833==--