Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:18 as permitted sender) client-ip=2620:137:e000::1:18;
Message-ID: <6275d3e7.1c69fb81.1d62.4504@mx.google.com>
Date:   Sat, 7 May 2022 02:05:25 +0000
From:   CGEL <cgel.zte@gmail.com>
To:     Michal Hocko <mhocko@suse.com>
Cc:     akpm@linux-foundation.org, hannes@cmpxchg.org, willy@infradead.org,
        shy828301@gmail.com, roman.gushchin@linux.dev, shakeelb@google.com,
        linmiaohe@huawei.com, william.kucharski@oracle.com,
        peterx@redhat.com, hughd@google.com, vbabka@suse.cz,
        songmuchun@bytedance.com, surenb@google.com,
        linux-kernel@vger.kernel.org, linux-mm@kvack.org,
        cgroups@vger.kernel.org, Yang Yang <yang.yang29@zte.com.cn>
Subject: Re: [PATCH] mm/memcg: support control THP behaviour in cgroup
References: <20220505033814.103256-1-xu.xin16@zte.com.cn>
 <YnUlntNFR4zeD+qa@dhcp22.suse.cz>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <YnUlntNFR4zeD+qa@dhcp22.suse.cz>
Precedence: bulk

On Fri, May 06, 2022 at 03:41:50PM +0200, Michal Hocko wrote:
> On Thu 05-05-22 03:38:15, cgel.zte@gmail.com wrote:
> > From: Yang Yang <yang.yang29@zte.com.cn> 
> > 
> > Using THP may promote the performance of memory, but increase memory
> > footprint. Applications may use madvise to decrease footprint, but
> > not all applications support using madvise, and it takes much costs
> > to re-code all the applications. And we notice container becomes more
> > and more popular to manage a set of tasks.
> 
> Could you be more specific about the actual usecase? When do you group
> processes based on their general THP reqirements? You are mentioning
> containers but those are usually bags of different processes that just
> share a common objective.
>
> > So add support for cgroup to control THP behaviour will provide much
> > convenience, administrator may only enable THP for important containers,
> > and disable it for other containers.
> 
> Why would that be a matter of importance?
> 
> Also what is actual semantic when processes living inside those cgroups
> explicitly state their THP requirements?
>
Docker might support this new cgroup knob in the future, add provide UI likes:
# docker run -it --thp-enabled=[always,never,madvise]
The cmdline format from https://docs.docker.com/engine/reference/run/

If there are many containers to run on one host, and some of them have high
performance requirements, administrator could turn on thp for them:
# docker run -it --thp-enabled=always
Then all the processes in those containers will always use thp.
While other containers turn off thp by:
# docker run -it --thp-enabled=never

By doing this we could promote important containers's performance with less
footprint of thp.
> -- 
> Michal Hocko
> SUSE Labs