New release with huge number of improvements, the most notable of which are
- column data storage (eblob only)
- range requests
- faster metadata and overall IO performance
- server-side checksumming and verification support
- base for ultra-fast recovery and local mapreduce
- simplified interfaces for POHMELFS
So far this is a beta release, since not all API calls were propagated to high-level interfaces. There are some issues with recovery process – we will not yet fully finished transition process. Eblob offline defragmentation tool was removed, but online processes was not yet added into the core.
We plan to resolve this issues this week and move further.
Elliptics currently hosts not-that-large clusters (the biggest cluster spawns over about 200 nodes which live on ~20 physical servers and close to 1 Pb of data) with about hundreds of millions objects (not counting redundant copies), which covers 4 datacenters including abroad. With above features we plan to expand our usage area. Unfortunately it is quite complex to get into with elliptics usage, and that should be my main goal actually.
Maybe eventually this will be a community driven project.
Anyway, stay tuned!