Kirk Pepperdine wrote:
>
> Jonathan,
>
> > an E10000 actually reduces total throughput, due to the extra context-switching
> > and/or memory contention overhead. We have a performance tuning paper which
> > illustrates this point.
>
> Can you share this white paper?
>
> Kirk
Kirk,
<vendor>
Unfortunately, we can only provide this document under NDA.
Why would this be? Well, this paper describes an implementation
of the TPC-W benchmark built using Session beans and CMP entity
beans on our AppServer. We implemented TPC-W, but did not go
through the process of auditing our implementation (which is a
very time consuming process). As per the TPC bylaws, unaudited
benchmark results can only be provided under NDA.
Interestingly, we had previously published this paper (not
realizing we had violated the bylaw), but we were threatened
with a legal action by BEA. Apparantly, BEA did not want
customers to know how our system scales. (Either that, or BEA
is surprisingly concerned about the TPC bylaws ;-)
Going forward, we have been involved with the ECperf benchmark.
Hopefully in the next few months, various vendors will start
to publish their ECperf-based performance numbers. This should
(finally) give customers a way to see which products have the
best performance (rather than which companies have the best-
performing legal departments). We will be publishing our ECperf
numbers as soon as possible. It would be good for users to
demand ECperf numbers from other vendors too.
At some point, potential customers will be able to download
ECperf configuration for each of their candidate AppServers,
and will very easily be able to see which runs best on their
given hardware setup.
</vendor>
-jkw
===========================================================================
To unsubscribe, send email to [EMAIL PROTECTED] and include in the body
of the message "signoff EJB-INTEREST". For general help, send email to
[EMAIL PROTECTED] and include in the body of the message "help".
