Supporting fast bucket aggregation on numeric multi field aggregation

[jainankitk]

Is your feature request related to a problem? Please describe

#9310 introduces approach for improving the date histogram aggregation without any sub aggregation. We can leverage similar approach to introduce new mapping type for storing multiple single dimensional as single multi dimensional field. It can then be used for optimizing bucket aggregation involving those fields.

If this is feasible, this should allow customers to significantly optimize latency of their common use case for slight indexing and storage overhead.

If this is feasible, this should allow customers to significantly optimize latency of their common use case for slight indexing and storage overhead.

Related component

Search:Performance

[peternied]

[Triage - attendees 1 2]
@jainankitk Thanks for filing this issue

[bowenlan-amzn]

We may research on this idea after supporting the sub aggregation #12602
Both targets are this operation from big5 workload

https://github.com/opensearch-project/opensearch-benchmark-workloads/blob/7c5ddaab4cba352d3a220d1ac7d571830b54fd57/big5/operations/default.json#L535-L581