user652241 wrote:Assuming you're talking about the benchmark that is discussed in my blog, we used YCSB as the load generator on the client side. It runs as a standard Java program (i.e. not in J2EE). We can start an arbitrary number of YCSB processes and each one can have any number of client threads performing operations.
For the benchmark you ran, was the code that went against the NoSQL DB an enterprise Java application? i.e., was it running in an application server cluster? Or, if not, how did you handle so many incoming requests?