I am running series of tests with and with out clustering. i am running an application that application is built on couchbase 2.2. I am running this on RHEL 6.3. With single node I get a average 2 ms response time (this is couch base activity + our application specific logic). For the same load when I add multiple node and even when I disable replication the response time increases to average 10 ms. These tests are run with same data set and same through put. When i look at the couch base web UI console the RAM usage is only 301 MB out of the 15.6 GB in total. Both the servers still have enough available memory. there is no disk fetch activity observed.
Appreciate your help in any tuning assistance and understanding what is the bottleneck here.