"Is it fair to expect that if one instance provides x requests/sec than 4 instances should provide 4x req/sec."
No, There is always some overhead involved in running a cluster. The only way to determine exactly what your scalability factor is, you would have to run load tests with first 1, then 2, then 3, and so on, nodes in the cluster. Plotting the various statistics should give you a good idea of the scalability of the cluster. Usually if it 90%, you are doing exceptionally well. By the way, with 90% scalability, a 4 node cluster would provide 3.6 times as much throughput as a single node. But usually you get around 70-80% scalabiltiy (4-node cluster is 2.8 to 3.2 times throughput).
Did you try running 2 and 3 nodes? It could also be that you have a bottleneck somewhere which you will have to find and resolve. Such performance tuning often takes a specialist. If you are not comfortable doing it yourself, or don't know how, you will have to hire someone to do it.