WhichPersistenceFramework

So SPerformance needs to store performance results for each microbenchmark. This should be tracked historically for comparison. It also needs the ability to efficiently identify clusters of data to query and do analytics.

The goal is to have a ClusterInfo block that the clustering pipeline can generate about data that includes some kind of efficient retrieval mechanism. This is combined with a raw-store that stores the original recorded values, so we don't overrun memory.

Other Objectives:

• Usable from any build system, with no setup
• Able to share data across machines and store in some permanent location.
• Very efficient retrieval based on desired clusters.

Current ideas for where to store data:

• Sqllite
• Derby database

Any ideas?

From Jesse: I have an XML format because I want portability. I think a DB is ok provided it can be downloaded from a remote location and easily read. I am guessing that makes Derby a simpler candidate? I have not used it too much I know H2 fits the bill as well but the fact that Derby is often installed along side java makes it pretty attractive.