Re: [Exist-open] JVM allocation

[Kemp, C. <ek...@ri...>]

Andrew, Jason and Loren,

Thank you all for your replies. Given a little bit more lead time we would have been able to implement the load balancing architecture described by Loren, and perhaps employ caching as Andrew suggests as well.

Jason, we're running 1.4.0 with Java version 1.6.0_17 and Jetty in production. For now we're doing some stress testing with Jmeter and will compare data to identical testing done previously. We'll then up the RAM allocation - probably to 4GB to the server and 2GB to JVM - increase the number of allocated CPUs, run our tests again and cross our fingers for the weekend.

Our intention is to move this application to a physical server in the near future - unfortunately we haven't had the staff time to get that done quite yet. When we do get around to the move, your responses have given us some substantial food for thought.

Best to all,
Chris

On Apr 13, 2011, at 12:17 PM, Andrew Welch wrote:

On 13 April 2011 16:32, Kemp, Chris <ek...@ri...> wrote:
Greetings,

We have eXist 1.4 running on a 32 bit Red Hat 5 virtual server with 1 GB RAM and a single 2.8 GHz CPU.

Our application is going to be featured in the NY Times this weekend in a column written by our university president, Edward Ayers. In anticipation of a spike in traffic we're adding resources to the virtual server that holds the eXist instance feeding the site (http://collections.richmond.edu/secession).

We have a couple of questions for the list:


 1.  Can a 32 bit Red Hat 5 JVM use more than 2 GB of memory?
 2.  If so, what is the upper limit of memory usage?

People tell me in situations like this its all about the caching...
check out Squid:  http://www.squid-cache.org/



--
Andrew Welch
http://andrewjwelch.com