Re:Looking with Cassandra into the future of Atlas
by jemfinch on Monday July 16, 2012 @07:43PM
(#4253296)
Sorry, I didn't mean to imply that Google or Amazon engineers implemented Cassandra (though it's notable, of course, that it was initially implemented by Facebook, adding a third Internet behemoth to its pedigree).
What I really meant to say is that it's clear that the engineers behind Cassandra have done their research and chosen an extremely well tested design, while the engineers behind MongoDB seem to be completely winging it, ignorant of the literature and writing (based on my last examination) extremely amateurish code.
That said, part of the reason Cassandra was attractive to me from the beginning is that unlike master-slave designs like MongoDB (or Bigtable/HBase, for that matter), a p2p design doesn't have the many corner cases around failover and recovery that complicate troubleshooting so badly. This is a primary reason Cassandra has had a very good reliability story since very early on.
[1] http://www.datastax.com/cassandrausers