This meeting is a hybrid teleconference and IRC chat. Anyone is welcome to join...here's the info:
- Review current fcrepo4 work
- RDF iteration in the core
- Simplified deployment
- Islandora interoperability
- Greg Jansen to check into possible cluster testing resource at UNC.
Review current work for Fedora 4
- RDF iteration in the core. (Refactor of code, move away from models and towards an lazy iteration model)
- Improved memory usage.
- Apache Jena, stored RDF in memory, contained more than we needed, now using our own abstraction.
- Simplified Deployment. (Osman Din, not present) Simplify configuration. Still in progress.
- Islandora Interoperability - Soon both Fedora 3 and Fedora 4 will share the same code base.
- Search (Ye Cao, not present), Generic Search Solution for Fedora 4. Mulgara to be built upon existing indexer (Nigel Banks).
Fedora Profiling (frank asseg)
- Some results from the SCC cluster: https://wiki.duraspace.org/display/FF/Performance+evaluation+on+the+SCC+Cluster
- We are currently limited by our test platform.
- Currently working on getting physical nodes to test the platform on rather than shared VM's.
- Fedora 4 now has a greater throughput than Fedora 3, yeah!
- Increasing the number of threads didn't seem to improve performance.
- Tweaking configuration.
- Check out project on Github https://github.com/futures/benchtool
- Single Node performance is important.
- LevelDB NoSQL store may work with infinispan to dramatically increase performance (Chris Beer tested this in single node before, may not be able to use with cluster).
- Two kinds of metrics:
- Load testing, this is what frank asseg is working on.
- Internal Metrics which time operations inside the software, good for finding the source of problems.