I'm evaluating some database possibilities for a large-scale (many billions of entries, many terabytes of data) storage solution where we will do random primary key-lookups almost exclusively.
Given it's capabilities, Membase (Couchbase 1.8) looks almost like a perfect fit, and some previous tests makes us believe it is highly performant for our usecase. Our main concern with using this though, is that since Couchbase 2.0 looks like a whole new direction from 1.8, the characteristics of the product might change. We like Couchbase 1.8 because it does exactly what we need very well.
We don't need views or map/reduce capabilities. While these are nice features to have, they are not something we need and not something we want if they are at all detrimental to performance. We've ruled out CouchDB mostly due to the complexity in scaling (adding/removing nodes) which is of course much better in Couchbase, but also because we had some concerns about it's disk usage.
Is anyone aware of any performance measurements made for 1.8 vs 2.0, disregarding all view and M/R capabilities?
Will the 1.8 fork continue to be maintained? Or is 1.8 dead, and we should just move on?
There will be no major differences in performance between Couchbase 1.8 and 2.0. I am a Couchbase employee and one of the most important things to us was that existing customers are able to upgrade from Couchbase 1.8 to Couchbase 2.0 whether or not views were important to them and have their applications continue to perform at the same levels.
I recommend starting with Couchbase 1.8 since our 2.0 product will not be released until the end of October 2012, but when the time comes to upgrade to 2.0 you shouldn't have any issues.
NOTE: As December 2012, Couchbase 2.0 is already available