[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <4524B0E9.8010005@web.de>
Date: Thu, 05 Oct 2006 09:14:49 +0200
From: Markus Wenke <M.Wenke@....de>
To: Alan Cox <alan@...rguk.ukuu.org.uk>
CC: linux-kernel@...r.kernel.org
Subject: Re: to many sockets ?
Alan Cox schrieb:
> Ar Mer, 2006-10-04 am 17:03 +0200, ysgrifennodd Markus Wenke:
>
>> Hi,
>>
>> I wrote a program which handles incomming sockets asynchron.
>> It can handle up to 140000 connections simultaneously while every
>> connection send some bytes in both directions continuously.
>>
>
> Armwavingly 64K x 2 per socket worst case for non tcp windowed buffering
>
> 128K per socket x 140000 connections
>
> 8750MB of RAM
>
> plus other overhead
>
> Assuming you kept the socket buffer limit to 64K by setting it or
> disabling window scaling you'd want a about 10GB of RAM for the sockets,
> buffering and resources. With tcp windows you'd need more.
The default values of my system are:
SO_SNDBUF = 16384
SO_RCVBUF = 87380
> If your data rates are always low, or the link is low latency you could
> set the send/receive socket buffer for each connection via setsockopt
> down to say 8K and come out needing perhaps 1GB or so instead.
>
I tried the same scenario with SO_SNDBUF = SO_RCVBUF = 8k, so that the
max memory is ca. 2G
and the oom-killer kills my application at the same time (at 140000
connections).
I can not see in the messages that the system is out of memory,
there is also no swap space used
You can download my /var/log/messages at
http://hemaho.mine.nu/~biber/messages
May you can give me a hint which line/value in the log shows me,
that the system is out of memory?
Thanks in advance
Markus Wenke
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists