-
Notifications
You must be signed in to change notification settings - Fork 1.9k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[Feature Request] Concurrent executing query in InnerHitsPhase #16878
Comments
Thanks for the feature request @kkewwei! Concurrent fetch is something I had briefly discussed with @sohami before, but it's not something we saw any use cases for (until now). When we were looking at concurrent search the performance on aggregations is what we were focusing on, and in a lot of cases users will do aggregations with either I'd be happy to review any designs or PRs for this! |
@jed326, If you haven't started on it, can you assign it to me? I'm pleased to have a try. Of course, I will make a draft first |
@kkewwei done! |
It seems that serial execution of each document's innerhit can speed up due to querycache, while concurrent executing will result in performance degradation. I'm still testing. |
Is your feature request related to a problem? Please describe
In our product environment, we utilize the
nested field
andinner_hits
, thesource
contains excessive number of fields, andinner_hits
query will match about 80 sub documents within a parent document. we find that the FetchPhase is costly, it will cost 3s+, but when we delete theinner_hits
, it will only cost 700ms.In
FetchPhase
, each document will execute the InnerHitsPhase serially (regards assub query phase
), if we need tofetch values from multi documents in
FetchPahse
, then the overall fetch phase will be very slow.Describe the solution you'd like
Concurrent executing
inner_hits
in FetchPhase.In some case, if the source is too large, it will also cost much time to fetch document in FetchPhase, It also seem necessary to fetch doc concurrently in some scenarios.
Related component
Search:Performance
Describe alternatives you've considered
No response
Additional context
No response
The text was updated successfully, but these errors were encountered: