- From: Yves Lafon <ylafon@w3.org>
- Date: Fri, 4 Dec 1998 11:53:43 +0100 (MET)
- To: Henry Story <henry.story@altavista.digital.com>
- cc: www-jigsaw@w3.org
On Thu, 3 Dec 1998, Henry Story wrote:
> Hello,
>
> I have been using the Java Web Server to develop a servlet.
> It is lacking some functionality I need, and so was thinking of moving
> to Jigsaw as the source code is available.
>
> I was wondering if Jisgsaw is inherently much slower that JWS?
Well, I didn't ran any performance comparison with JWS, the only test done
are between Apache and Jigsaw, and Jigsaw is not much slower (using the
optimized jdk for solaris). It mainly depend on the modelized traffic.
> Also I looked through the archive of the Jigsaw mailing list and
> was surprised to see that it is so quiet. There are not that many posts
> on usenet on the subject either. Am I looking in the wrong place?
No, but the majority of the posts are not on the ML, unfortunately!
> The service I am thinking of using will easily get 1million hits
> a day. I was going to compile it with TowerJ to speed the server up.
> Any thoughts on this either?
1Mhit/day ~ 12 hits/sec, which is far less than Jigsaw can handle. Of
course it depends on the processing you do behind the scene.
I know that TowerJ has been used successfuly to compile Jigsaw, and that
it should reduce a lot the impact on the CPU. Unfortunately I did not test
this compiled version, so I can't give any number.
But I am pretty sure that a compiled version should handle that even with
processing behind.
Note that the latest available sources can be found from the CVS tree at:
http://dev.w3.org/cgi-bin/cvsweb/java/
Regards,
/\ - Yves Lafon - World Wide Web Consortium -
/\ / \ Architecture Domain - Jigsaw Activity Leader
/ \ \/\
/ \ / \ http://www.w3.org/People/Lafon - ylafon@w3.org
Received on Friday, 4 December 1998 06:04:17 UTC