[Dspace-general] High performance Dspace

[Sam Ottenhoff]

> I would like some information on how 
> Dspace can be modified for high performance, scalability and redundancy. 
> We are thinking of a site that is robust enough to can handle 500 
> simultaneous users. 

Hi Yannick,

If these 500 simultaneous users are non-authenticated, browsing users, 
you should probably look to reduce the strain on the Tomcat application 
server and the database by using a reverse caching proxy.  Squid 
(www.squid-cache.org) is the most common reverse proxy and can cache the 
web content in memory and on disk.  If the caching proxy already has the 
web content in memory (and it is fresh), it will simply deliver the 
request without invoking DSpace. Two high-performance reverse-proxy 
servers with large amounts of RAM should be able to handle five hundred 
non-authenticated, browsing users without a problem.

If these 500 simultaneous users are all authenticated users, you will 
need to explore Tomcat clustering, load balancing (specific hardware or 
via Apache), and database clustering.  This DSpace wiki page is a start: 
http://wiki.dspace.org/index.php/Clustering

--Sam