As part of performance improvement we are caching the objects in memory . Once an object is modified we are invalidating the cache using JMS .
But because of the asynchronous nature their is some latency in invalidating the cache after an update and the same objects are getting read by different requests during this time (which is actually dirty data...).
Any suggestions in over coming this issue or any better design ?
If you are looking for a piece of third-party software to avoid this issue Tangosol's Coherence provides all the facilities to cache data in a cluster. This is done coherently and synchronously to avoid 'dirty-data' and 'missing-updates', complete with cluster-wide locking and event notification.
Coherence: Easily share live data across a cluster!