On Sat, May 14, 2011 at 07:39:30PM +0200, Eugen Leitl wrote:
> How braindead are the current official squeeze kernels?
> Can they be used in production or is there show-stopping
I use these kernels on multiple servers, they work for my usecases:
isolation of different services like file, mail, dns, web, sql-servers.
However, recently I have the feeling servers with a few VServers and
many processes behave bad in IO. Probably this is true for non-VServers
as well, but I tend to not have these for verification.
In numbers: Adaptec 5408 controller, does >300MB/s with the attached
disks, with 1300+ processes on the 'idle' server, iotop shows ~5-15MB/s
during the day and load 4-20, depending on workhours. One of the
processes is a backup client backing up big files.
As said, it is a feeling, I have no laboratory benchmark numbers and as
I use 'DRBD' on these servers as well and DRBD-IO doesn't show in iotop
it might be completly different.
Actually none of the users of the server complained about a slow server,
but getting a new shell takes its time.
This is the background for my cgroup question earlier today, I
understand they change the scheduler to schedule groups and might be
this improoves my performance experience.
Be aware, my VServers are potentially friendly towards each other, I use
them for easier management, I don't do tests if the separation actually
works or has any flaws.
-- LiHAS - Adrian Reyer - Hessenwiesenstraße 10 - D-70565 Stuttgart Fon: +49 (7 11) 78 28 50 90 - Fax: +49 (7 11) 78 28 50 91 Mail: lihas_at_lihas.de - Web: http://lihas.de Linux, Netzwerke, Consulting & Support - USt-ID: DE 227 816 626 StuttgartReceived on Sat May 14 20:12:37 2011