Starting from 3, maybe 4 nodes, I would prefer using Memcached.
+1 for memcached in this scenario. In my refactoring work on the memcached storage backend I removed the async get/set methods to avoid this situation. I would argue all CAS storage backends should take into account state replication delays to prevent this kind of behavior. While a failure rate of 0.3% is probably acceptable to most deployments, we should strive for 0 by design.
So this will "just work" with memcached, whereas you may need to do some careful tuning to get it to work 100% of the time with a replicated Ehcache solution.
M -- You are currently subscribed to [email protected] as: [email protected] To unsubscribe, change settings or access archives, see http://www.ja-sig.org/wiki/display/JSG/cas-user
