For me, the take away from this article is that MemSQL's ACID compliance is half-assed at best, which is a much more damning criticism than using a flawed benchmarking methodology.