The numbers were suspect because I was running 3.3 under
JBuilder but 4.0 externally, I'd imagine. I set up clean
tomcat trees and ran them outside of any test environment
and I get these numbers: (100 rqs/10 concurrent cx)
TC 4.0.6 | TC 3.3.2
Http11 Http10 | Http11 Http10
|
Rq/sec 26.9 25.89 | 18 27
kb/sec 393 377 | 266 404
Coyote/11 with Tomcat 3.3 is still a dog. I would imagine
you'd get the similar numbers with any servlet as I changed
nothing but the servlet container between the tests.
I'm still investigating, but it may be easier for us to
move to 4.0.
Keith
| -----Original Message-----
| From: Remy Maucherat [mailto:remm@;apache.org]
| Sent: Thursday, November 07, 2002 2:22 AM
| To: Tomcat Developers List
| Subject: Re: Coyote/Http11 under load
|
|
| Keith Wannamaker wrote:
|
| > Same box, same app, same test, with Tomcat 4.0.6's http10 and 11 adaptor:
| >
| > coyote/http11 37rq/sec, 406kb/sec
| > http10 30rq/sec, 181kb/sec
| >
| > I've always held that 3.3 was more "lean and mean" than 4.0 but
| > now I'm wondering if this is just indicating some basic architectural
| > differences.
|
| I don't understand those numbers. Even the two above (how can there be
| twice the bandwidth with only 20% more requests if the bench is correct ?).
| I thought 3.3 + Coyote HTTP/1.1 was as fast as the default HTTP/1.0.
|
| Remy
|
| >
| > Keith
| >
| > | -----Original Message-----
| > | Subject: Coyote/Http11 under load
| > |
| > |
| > | I'm using http11 on a busy site with more or less Tomcat 3.3 head
| > | (3.3.2-dev).
| > |
| > | coyote/http11 4.2rq/sec, 63kb/sec
| > | http10 6.3rq/sec, 92kb/sec
|
|
--
To unsubscribe, e-mail: <mailto:tomcat-dev-unsubscribe@;jakarta.apache.org>
For additional commands, e-mail: <mailto:tomcat-dev-help@;jakarta.apache.org>
