[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CABWYdi2L_qp8SmZ_w3pSSracYHEVku3TaBoXL7E0Nzn7CN3neg@mail.gmail.com>
Date: Fri, 14 Jul 2023 17:14:19 -0700
From: Ivan Babrou <ivan@...udflare.com>
To: Shakeel Butt <shakeelb@...gle.com>
Cc: cgroups@...r.kernel.org, Linux MM <linux-mm@...ck.org>,
kernel-team <kernel-team@...udflare.com>,
Johannes Weiner <hannes@...xchg.org>,
Michal Hocko <mhocko@...nel.org>,
Roman Gushchin <roman.gushchin@...ux.dev>,
Muchun Song <muchun.song@...ux.dev>,
Andrew Morton <akpm@...ux-foundation.org>,
linux-kernel <linux-kernel@...r.kernel.org>
Subject: Re: Expensive memory.stat + cpu.stat reads
On Mon, Jul 10, 2023 at 4:21 PM Ivan Babrou <ivan@...udflare.com> wrote:
> The fast one is v6.1.37 and the slow one is v6.1.25. I'm not sure if
> the kernel version makes a difference or if it's a matter of uptime /
> traffic profile. The data is from two different locations. The fast
> location has gone through an expansion, which meant a full reboot with
> a kernel upgrade, so maybe that affected things:
>
> * https://i.imgur.com/x8uyMaF.png
>
> Let me try to reboot the slow location and see if there's any lasting
> improvement.
There is no correlation with the kernel version, v6.1.38 is slow too:
completed: 23.09s [manual / cpu-stat + mem-stat]
completed: 0.30s [manual / mem-stat]
completed: 0.64s [manual / cpu-stat]
Powered by blists - more mailing lists