- A Fixed Memory Buffer - A serialization buffer will be allocated to increase disk through-put. It is also managed as a second tier cache. You may optionally configure the size and set it to use off-heap memory - which is especially useful for dealing with larger amounts (typically 32BG+) of memory from the VM. The memory buffer also allows for better disk usage based upon fixed allocation blocks vs. the old Teiid scheme which used a file per table/tuple buffer.
- Concurrent LRFU - the main eviction algorithm was further refined to be concurrent and to better handle table work loads.
- Numerous Performance Improvements - batch size estimation, serialization, and value type resolution were all improved.
Now is a great time to provide feedback on any issues or minor enhancements before it's too late for the release.
The Teiid Team