Announcing HyperLevelDB

[Robert Escriva]

We are proud to announce HyperLevelDB, a data store crafted especially
for HyperDex (http://hyperdex.org), our high-performance, consistent,
fault-tolerant, transactional data store. Derived from LevelDB,
originally from Google, HyperLevelDB provides:

(1) improved parallelism by using finer-grained locking to achieve
higher throughput for multiple writer threads, and
(2) improved compaction to achieve higher throughput for write-heavy
workloads, especially for large databases.

We've posted performance comparisons between LevelDB and HyperLevelDB on
the HyperDex website[1]. HyperLevelDB scales better than LevelDB with
concurrent threads, and achieves up to 4X higher throughput for
write-heavy workloads.

HyperLevelDB is the default data backend for the upcoming release of
HyperDex.

Happy Hacking,
The HyperDex Team

[1] http://hyperdex.org/performance/leveldb/