Elasticsearch / Marvel {search rate & search latency }

I had the same question before and posted on Marvel forum, but no one answered me.

Later I found it out by the observation that when the number of shards of an index changes the search rate changes as well. We only have one index for search right now, and when there is a single shard, the search rate is almost the same as the requests we send to ES cluster. When we have 10 shards, it’s 10 times the actual number of requests.

So, I believe:

1. Search Rate is actually the search requests happens on all shards per second
2. Search Latency is the latency within each shard