Hi every body, if I add a few servers to the cluster in replica role, will it improve cluster search performance? or I will get only data copy and cluster performance will have no changes.
There's no such thing as "replica roles" for nodes. All data nodes are equals. It's likely that you'll get a performance gain, but it depends on how much data you have, how it's sharded, what kind of hardware you have, what kind of queries you make, and so on.
You might want to look at the client node. Client nodes do not run the data service and only serve as a HTTP loadbalancer of sorts for the cluster.
Thanks for the link!