write through / read through caching
posted 9 years ago
I'd like to canvas peoples opinions on increasing the scalability of some software I'm working with, this is an EJB application using stateless session beans, bean managed persistence and a database as the hard store. One of the most performance critical elements of the software are a set of batch processes which at present process their work load on a job by job basis. We have identified that we could make this far more scalable if the EJB middle tier processed it's workloads in multiple jobs at a time and accessed the database in such a way that single queries could provide the data necessary for servicing multiple jobs. However, this approach would require a major re-write exercise. Although we are considering this, another approach would be to use some caching software such as Coherence or Giga Spaces application fabric to act as a read through / write through cache to the database. I was wondering if someone could provide me with the benefit of any experience they have when encountering a similar problem.
author and iconoclast