Hi, On 11/24/06, Stefan Gruszczynski <[EMAIL PROTECTED]> wrote:
So if not RMI than what? Did you or anybody else performed any performance tests not based on RMI?
I think the most frequently used deployment models for Jackrabbit are models 1 and 2 where the repository is running within the same JVM as the client application. There have been some performance tests but so far we've seen very little raw data or consistent reports being sent back to the community. Having your results available is very valuable!
I think that some of those tests are not affected by RMI disadvantages /i.e. deleting nodes or pasring XML documents/ because those operations are running on server side only - without or with small communication with client - what about cpu load and memory usage - does RMI affect it in any way?
You're correct, see my other mail for the most likely bottlenecks. The RMI layer does very little work by itself so it should have a negligible effect on CPU loads and shouldn't introduce too much memory overhead. BR, Jukka Zitting
