On Tue, Mar 13, 2012 at 10:58 AM, Aristotle Pagaltzis <pagalt...@gmx.de> wrote: > * David Golden <xda...@gmail.com> [2012-03-12 20:20]: >> MongoDB is the leading candidate to replace Amazon's SimpleDB > > I hope that works out if you try it. You have heard the stories? > <http://pastebin.com/raw.php?i=FD3xe6Jt> > <http://blog.schmichael.com/2011/11/05/failing-with-mongodb/>
Yes, I've heard the stories. Fortunately 10gen is here in NYC and I've been very impressed with their people and had some candid conversations about the stories as well. I've asked hard questions about some of the stories and talked to CTO's for a half dozen companies that use it about their positive/negative experiences. I haven't seen any real roadblocks. Like most things, if you know what you're doing and aren't stupid (e.g. trying to shard when you're already at 100% resource utilization, choosing the wrong shard key), it works pretty well. In my own testing vs sqlite and postgres, MongoDB scaled writes (single process) much better. In a sharded configuration, it should be lovely. > Have you evaluated Kyoto Cabinet? <http://fallabs.com/kyotocabinet/> It's not really sufficient because it's only key/value store (as I understand it). MongoDB gives a query interface that isn't SQL, but still is multi-dimensional. Plus it allows multiple indexes, which is really useful for us since people want queries on time (for recent reports) and queries on distname/version and we can make both cheap. The CT usage pattern is write-heavy, with reads mostly against the most recent reports. That's really good for keeping the relevant index/data pages in memory, which is ideal with MongoDB. -- David
