The COSMOS data collection framework is cognizant of the existence of existing data stores for monitored data. The key requirements for COSMOS are simplicity and a flexible object/relational mapping. When we started the initial work in COSMOS, for good or bad, we focused on execution and did not dig deeply into the pros or cons of using a particular persistence framework. As we move forward, the upcoming F2F in Toronto provides an opportunity to have this discussion and objectively evaluate our options.
This page is intended for us to have this discussion. It is a continuation of the e-mail thread started between Joel and Mark. The discussion on this page will be worked into the data collection architecture document being iterated on in CVS, and will then be made available on the COSMOS web site.
Here is a talk/discussion page on this.
COSMOS Persistence FW Core Requirements
There are only a handful of core requirements for the COSMOS persistence framework.
- It must be simple for the user, be easily adoptable, and have low barriers of entry. We do not want to inadvertently create a barrier of adoption for COSMOS.
- It must provide flexible handling of object <-> relational mapping.
Persistence Frameworks & Complexity
There are a number of persistence frameworks that satisfy the O/R requirement, but have varying degrees of complexity. The diagram below lists a number of the popular persistence frameworks along with their typical usage. It also places them along a complexity scale.
Current work in COSMOS