Doing it right (was Re: Preparations for ZD's upcoming Apache/Linux benchmark)

Raul Miller (rdm@test.legislate.com)
Tue, 8 Jun 1999 08:57:55 -0400

Linus Torvalds <torvalds@transmeta.com> wrote:
> However, the things is that web performance obviously _is_ important to
> a lot of people, and it's really hard to benchmark slow connections and
> lossy networks etc which is what most real-life uses are. So webbench
> is kind of a "default choice" - it's a bad benchmark of real-life
> behaviour, but there isn't anything else out there.

Hmm.. it should be possible to write a decent benchmark...

Basically, you'd want to have a machine (or a set of machines, if one
machine wouldn't be up to it) which takes over a whole range of addresses
(maybe the 10.*.*.* bogusnet) and did things on a packet-at-a-time basis.

You'd probably want to rip the tcp code out of the kernel and bring it
into user space where you could mutilate it.

You'd probably want something reasonably decent for your top level,
so you could change what's being accessed, and how.

However, I'm not enough of a web performance guru to suggest lotsa
details.

--
Raul


-
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/