SP2013 : Search Components design

In one of my previous article, I talked about Search topology configuration in SP2013. In this blog, I want to highlight on some of the key considerations for designing search architecture in a large environment. 

Although technically you can have multiple Index replicas in one server, it is highly recommended not to have more than one index replica per server. Say for e.g. you have 20 million items for which there will be 2 index partitions with 2 index replicas per partition for fault tolerance. Per recommendation you will need to design to have 4 search servers to host 4 index replicas in total ( 2 replicas/partition).

The same theory goes for Content Processing Component (CPC) – recommended not to have more than 1 CPC per server. The other search components(Query Processing Component, Crawl component, Analytics, Search Administration) can be hosted in the same server or distributed across different servers.