bench

> It seems as if it's designed for situations where you're doing a lot of table scans. This is not quite true. Lockfree data structure work exceptionally well for a lot of small concurrent reads and writes. If you look at the benchmarking code: https://github.com/memsql/bench/blob/master/sql_database.py you will notice that there are a lot of queries like "fetch by pk/guid". The trade-offs are: taking a hit the first time you run the query, being limited to the amount of RAM, limited set of features for this release. The benefits are throughput for "fetch by pk/guid" transactions and predictably low latencies.