Multi-index overhead

If I have a situation wherein there are a large number of aliases all pointing to a few concrete indices.
For example, 10,000 aliases, with each of them pointing to one or more of just 5 concrete indices.

In this situation, suppose I were to run a multi-index Search query on all the aliases.

Would there be a performance penalty in using the multi-index search query (consisting of comma-separated list of the 10,000 aliases) as opposed to using 5 concrete indices the multi-index query?

Thank you

A bit of overhead associated with resolving the aliases to their concrete indices, but when the search is executed it'll just be directed at the concrete indices. So the bulk of the time will be the same since most of the work is in the search request, not the alias resolution.

There is some overhead to aliases in the form of cluster state updates... more aliases == more items needed to be tracked in the cluster state. That can slow down updates by the masters, and is more information that needs to be published to the cluster.

So thousands of aliases is generally ok, hundreds of thousands/millions is starting to get abusive :slight_smile:

1 Like

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.