Lots of small indices is an anti-pattern due to performance overhead. I’ve seen such a setup wreck cluster stability. By sticking to shards that are 10-100GB,

Lots of small indices is an anti-pattern due to performance overhead. I’ve seen such a setup wreck cluster stability. By sticking to shards that are 10-100GB, and only using i3 instances, you can get great performance. My company ingests about 2 TB per day and we use less than 10 i3 instances.