We currently have a Resin cluster of 25 nodes, configured to use JDBC
persistence. We typically have around 140,000 active sessions.
I am considering moving to cluster store persistence. We have been
running on a single MySQL database. I have started to look into
clustring it and thoght maybe cluster store would be better.
Our app doesn't use the session for anything other than login state and
the cluster simply provides SSO and LB.
How would I go about determining what memory consumption our sessions
would add to each node's JVM? Is there any documentation that would show
how much memory 1 session would take in it's basic form? I could then
extrapilate that out to our load.
Any perf gain/loss by going to cluster store?
My goal is better HA of session data and no performance degradation.
resin-interest mailing list