[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <8a7c05f2-671a-7540-40f5-ac8809a2f869@google.com>
Date: Wed, 9 Jul 2025 12:14:25 -0700 (PDT)
From: David Rientjes <rientjes@...gle.com>
To: "Christoph Lameter (Ampere)" <cl@...two.org>
cc: Kent Overstreet <kent.overstreet@...ux.dev>,
Casey Chen <cachen@...estorage.com>,
Andrew Morton <akpm@...ux-foundation.org>, surenb@...gle.com,
corbet@....net, dennis@...nel.org, tj@...nel.org,
Vlastimil Babka <vbabka@...e.cz>, mhocko@...e.com, jackmanb@...gle.com,
hannes@...xchg.org, ziy@...dia.com, roman.gushchin@...ux.dev,
harry.yoo@...cle.com, linux-mm@...ck.org, linux-kernel@...r.kernel.org,
linux-doc@...r.kernel.org, yzhong@...estorage.com,
Sourav Panda <souravpanda@...gle.com>
Subject: Re: [PATCH] alloc_tag: add per-NUMA node stats
On Tue, 8 Jul 2025, Christoph Lameter (Ampere) wrote:
> > Right, per-node memory attribution, or per zone, is very useful.
>
> I thought that was already possible using cgroups? That way you have the
> numbers for an application which may be much more useful.
>
>
Cgroups can tell us the nature of the memory if it's charged, yes. Memory
Allocation Profiling, which this patch series extends, however, provides
much more detailed insight such as the caller that allocated the memory.
It's been invaluable to identify memory efficiency savings and now, with
Casey's patch, can be used to identify where NUMA imbalances exist for the
same callers.
Powered by blists - more mailing lists