Hi
I have various of vector if the number of data increase I will face the problem of ram how to solve it?
Can we search in two collection for search for one input?

Hi!
There are several different solutions:

• use mmaps, in this case only a fraction of data will be stored in ram. It might impact the search speed, though
• use horizontal scaling and distribute the collections across multiple shards.

i have the same problem, for me the solution was deploy several servers but only if you have the availability of increase the servers, that keep the precision and speed (depend in the lantency of the servers)

Disk IOPS have a significant role in the performance, if you are going to into mmap option. And cloud providers tend to give a very limited performance on disks by default. Especially it is bad if it is a network-mounted disk