MemSQL 3.1 发布了，值得关注的改进有：
Unmaterialized views which enable row- and column-level security with online ALTER VIEW. Automatic
cross-datacenter replication via a single REPLICATE DATABASE command.
The slave site can be used as a hot backup and for offloading read
processing from the main site. INSERT and LOAD DATA are now transactional in the distributed system. Both commands now also support the IGNORE syntax. INSERT...SELECT
in the distributed system. INSERT...SELECT will automatically push down
to the leaves across a foreign shard key or stream through the
aggregator otherwise. The aggregator can now stream GROUP
BY queries rather than materializing a hash table. This has major
performance and space implications for high cardinality GROUP BY
queries.
由前Facebook工程师创办的MemSQL，号称世界上最快的分布式关系型数据库，兼容MySQL但快30倍，能实现每秒150万次事务。原理是仅用内存并将SQL预编译为C++。
MemSQL 提供免费的开发者版本（数据限制32G）和全功能试用版本（试用期30天） | 
