Re: any documentation on sun's reference RMI vs weblogic's RMI

On Wednesday 03 May 2006 22:51, Robert Walker wrote:
> anyhow, I am wondering at which point (if ever)
> do we need to consider other implementations of RMI
> and if anyone has some documentation on
> load comparisons.

I don't, but looking at the network nature and basic experience with scale-up
on RMI... We have found that the most limiting factor is revolving around
"calls-per-second". With average calls, this figure seems tightly coupled to
the latency of TCP/IP combined with the synchronicity of RMI.
The second general limiter for us has been general serialization speed,
although CPU improvements over the years have made this point a lot more moot
than it used to be.

My guess is that your system chokes at somewhere around 300-500 calls per
second per switched network card, due to RMI and with the hardware you are
talking about you should have only nudged the CPU cycles a little bit with
that, leaving a majority for your business logic.

The obvious way to improve performance (soft-wise) is to make less number of
calls, which often is possible to introduce without any client changes. You
could introduce a smart proxy which uses a different server interface which
allows you to transfer more information in one go, buffering it locally and
exposing out the same old interface to the rest of the client application.


Rickard Oberg's book "Mastering RMI" have a chapter discussing scalability,
that could be providing additional info.


Cheers
Niclas

===========================================================================
To unsubscribe, send email to [email protected] and include in the body
of the message "signoff RMI-USERS".  For general help, send email to
[email protected] and include in the body of the message "help".

For a list of frequently asked RMI questions please refer to:
http://java.sun.com/j2se/1.3/docs/guide/rmi/faq.html

To view past RMI-USERS postings, please see:
http://archives.java.sun.com/archives/rmi-users.html