That not the point of the article, which is not very well made. The author explains/try to explain that MVCC has also its quirks under high write workload, it's

That not the point of the article, which is not very well made. The author explains/try to explain that MVCC has also its quirks under high write workload, it's not a silver bullet, even when considering "eventual consistency databases" which "do nothing to help you", quite the contrary because it's a design choice. I think this article must be read inside the debate "ACID or no ACID". Where Google with F1 [1] and appengine [2], FoundationDB and wiredtiger push for strong transaction features because it's easier for developers. [1] http://research.google.com/pubs/pub38125.html http://research.google.com/pubs/pub38125.html [2] A Design for a Distributed Transaction Layer for Google App Engine