Re: Very poor TCP/SACK performance

Savochkin Andrey Vladimirovich (saw@msu.ru)
Tue, 8 Sep 1998 16:28:08 +0400


On Tue, Sep 08, 1998 at 02:13:48PM +0100, Alan Cox wrote:
> > The problem is, with the default settings we are pretty safe about
> > buffer overcommit. It also makes "I'm going to send you out of order
>
> In 2.0.x I used 100% overcommit. With ethernet drivers doing rx_copybreak
> that comes out about right in the wash. It was originally a finger in air
> and gues exercise but it turned out near enough right

I'd suggest to increase the default receive buffer limit to about
128K (admins who bother about it could tune it) and keep the send buffer
limit as it is.

I suppose that per user resource accounting can help admins to fight
against overcommits more efficiently.
I'm working under per user resource accounting in the background :-)
The accounting of page directory space is close to be finished.
Socket buffer space is the next in the todo list.

Best wishes
Andrey V.
Savochkin

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/faq.html