memcache should be more faster than mysql or postgresql in retriving and inserting since it is in mem! This comparision and benchmarks are invalid.

can i get the schema that is used for performance testing.

What is unclear (and I'd be happy to see more material) is what type of data was compared.

SQL is going to be a lot faster when (a) you can insert an entire document with one INSERT instead of multiple inserts for denormalized tables, and (b) you can select from one table, instead of using JOINs. Neither the chart above nor the PDF tell me the structure of the data tested, and benchmark-wise, that's where one should hypothetically see the performance difference between document DBs and RDBMS systems.

I, too, would also rather see comparisons between C-level drivers. That will seriously stunt a speed comparison if one is using a C library and the other is using an interpreted language.

the "Download OpenSQL comparison PDF" link is wrong...

Thanks! Fixed now!

Thanks for the writeup. One thing to note about the above benchmark is that the tests were performed using the Ruby driver from a single client *without* the C extension, which greatly improves performance. If the C extension were used (or one of the faster drivers, like Java) you'd see greatly better numbers for MongoDB. The limiting factor in this test was the client, not the MongoDB server itself.