Hi @kporter
Thanks for your replies; they are so much helpful 
Your points on service/transaction threads were correct and modifying them has given me better results. But rather than modifying them directly, I used the afterburner script provided by AS and it tuned those configs itself. I hope that’s a correct way as well.
My new result is 1.4M TPS and this is really good. But I have hit another hurdle after continuing further with my experiments.
This time I have created a cluster of two nodes(earlier was just 1 node) and a replication factor of two. But rather than improving, my TPS actually falls to 0.9 M TPS. This is very strange as we of-course expected it to increase.
Another interesting point is that our AS server maxed out with CPU usage percentage of 70-75% and load average of 25-30. So, bottleneck was AS, not the network. I was also monitoring n/w usage through netload tool and it was not hitting max capacity. No, of clients I am using is 6(host)*120(threads).
I have seen this thread and it suggest migration and replication as possible problems. In my case, migration is not an issue as both my nodes start together on a clean slate. But replication is something which we want alongwith clustering in production and will like to work it as expected.
Kindly help on this as how we can achieve better TPS through clustering and replication. Thanks.