ESQL: Added LIMIT BY benchmarks

[ivancea]

Draft

To be merged after LIMIT BY feature is released (In tech preview): elastic/elasticsearch#112918

Added benchmarks similar to the existing "Large Schemas" (LIMIT) and TopN (SORT + LIMIT) ones.

Local benchmark results

LIMIT BY

Metric	Task	Value	Unit
Min Throughput	from_idx_limit_1_by_key_1000	4.31	ops/s
Mean Throughput	from_idx_limit_1_by_key_1000	9.63	ops/s
Median Throughput	from_idx_limit_1_by_key_1000	10.34	ops/s
Max Throughput	from_idx_limit_1_by_key_1000	13.36	ops/s
50th percentile latency	from_idx_limit_1_by_key_1000	48.3455	ms
90th percentile latency	from_idx_limit_1_by_key_1000	62.2281	ms
99th percentile latency	from_idx_limit_1_by_key_1000	85.7927	ms
100th percentile latency	from_idx_limit_1_by_key_1000	119.196	ms
50th percentile service time	from_idx_limit_1_by_key_1000	48.3455	ms
90th percentile service time	from_idx_limit_1_by_key_1000	62.2281	ms
99th percentile service time	from_idx_limit_1_by_key_1000	85.7927	ms
100th percentile service time	from_idx_limit_1_by_key_1000	119.196	ms
error rate	from_idx_limit_1_by_key_1000	0	%
Min Throughput	from_idx_limit_10_by_key_1000	13.3	ops/s
Mean Throughput	from_idx_limit_10_by_key_1000	16.82	ops/s
Median Throughput	from_idx_limit_10_by_key_1000	17.58	ops/s
Max Throughput	from_idx_limit_10_by_key_1000	18.23	ops/s
50th percentile latency	from_idx_limit_10_by_key_1000	48.3263	ms
90th percentile latency	from_idx_limit_10_by_key_1000	55.2346	ms
99th percentile latency	from_idx_limit_10_by_key_1000	76.0557	ms
100th percentile latency	from_idx_limit_10_by_key_1000	158.521	ms
50th percentile service time	from_idx_limit_10_by_key_1000	48.3263	ms
90th percentile service time	from_idx_limit_10_by_key_1000	55.2346	ms
99th percentile service time	from_idx_limit_10_by_key_1000	76.0557	ms
100th percentile service time	from_idx_limit_10_by_key_1000	158.521	ms
error rate	from_idx_limit_10_by_key_1000	0	%
Min Throughput	from_idx_limit_1000_by_key_1000	9.05	ops/s
Mean Throughput	from_idx_limit_1000_by_key_1000	9.2	ops/s
Median Throughput	from_idx_limit_1000_by_key_1000	9.19	ops/s
Max Throughput	from_idx_limit_1000_by_key_1000	9.37	ops/s
50th percentile latency	from_idx_limit_1000_by_key_1000	99.4275	ms
90th percentile latency	from_idx_limit_1000_by_key_1000	119.911	ms
99th percentile latency	from_idx_limit_1000_by_key_1000	176.433	ms
100th percentile latency	from_idx_limit_1000_by_key_1000	225.489	ms
50th percentile service time	from_idx_limit_1000_by_key_1000	99.4275	ms
90th percentile service time	from_idx_limit_1000_by_key_1000	119.911	ms
99th percentile service time	from_idx_limit_1000_by_key_1000	176.433	ms
100th percentile service time	from_idx_limit_1000_by_key_1000	225.489	ms
error rate	from_idx_limit_1000_by_key_1000	0	%
Min Throughput	from_idx_limit_10000_by_key_1000	10.22	ops/s
Mean Throughput	from_idx_limit_10000_by_key_1000	10.42	ops/s
Median Throughput	from_idx_limit_10000_by_key_1000	10.34	ops/s
Max Throughput	from_idx_limit_10000_by_key_1000	10.74	ops/s
50th percentile latency	from_idx_limit_10000_by_key_1000	89.1841	ms
90th percentile latency	from_idx_limit_10000_by_key_1000	101.843	ms
99th percentile latency	from_idx_limit_10000_by_key_1000	136.421	ms
100th percentile latency	from_idx_limit_10000_by_key_1000	141.817	ms
50th percentile service time	from_idx_limit_10000_by_key_1000	89.1841	ms
90th percentile service time	from_idx_limit_10000_by_key_1000	101.843	ms
99th percentile service time	from_idx_limit_10000_by_key_1000	136.421	ms
100th percentile service time	from_idx_limit_10000_by_key_1000	141.817	ms
error rate	from_idx_limit_10000_by_key_1000	0	%

SORT + LIMIT BY

Metric	Task	Value	Unit
Min Throughput	topnby_by_payment_type_query	0.1	ops/s
Mean Throughput	topnby_by_payment_type_query	0.1	ops/s
Median Throughput	topnby_by_payment_type_query	0.1	ops/s
Max Throughput	topnby_by_payment_type_query	0.1	ops/s
50th percentile latency	topnby_by_payment_type_query	9815.24	ms
90th percentile latency	topnby_by_payment_type_query	9981.46	ms
100th percentile latency	topnby_by_payment_type_query	10060.8	ms
50th percentile service time	topnby_by_payment_type_query	9815.24	ms
90th percentile service time	topnby_by_payment_type_query	9981.46	ms
100th percentile service time	topnby_by_payment_type_query	10060.8	ms
error rate	topnby_by_payment_type_query	0	%
Min Throughput	topnby_by_payment_type_esql_segment_partitioning	0.21	ops/s
Mean Throughput	topnby_by_payment_type_esql_segment_partitioning	0.22	ops/s
Median Throughput	topnby_by_payment_type_esql_segment_partitioning	0.22	ops/s
Max Throughput	topnby_by_payment_type_esql_segment_partitioning	0.22	ops/s
50th percentile latency	topnby_by_payment_type_esql_segment_partitioning	4569.03	ms
90th percentile latency	topnby_by_payment_type_esql_segment_partitioning	4674.38	ms
100th percentile latency	topnby_by_payment_type_esql_segment_partitioning	4867.54	ms
50th percentile service time	topnby_by_payment_type_esql_segment_partitioning	4569.03	ms
90th percentile service time	topnby_by_payment_type_esql_segment_partitioning	4674.38	ms
100th percentile service time	topnby_by_payment_type_esql_segment_partitioning	4867.54	ms
error rate	topnby_by_payment_type_esql_segment_partitioning	0	%
Min Throughput	topnby_by_payment_type_esql_doc_partitioning	0.21	ops/s
Mean Throughput	topnby_by_payment_type_esql_doc_partitioning	0.21	ops/s
Median Throughput	topnby_by_payment_type_esql_doc_partitioning	0.21	ops/s
Max Throughput	topnby_by_payment_type_esql_doc_partitioning	0.21	ops/s
50th percentile latency	topnby_by_payment_type_esql_doc_partitioning	4651.93	ms
90th percentile latency	topnby_by_payment_type_esql_doc_partitioning	5207.24	ms
100th percentile latency	topnby_by_payment_type_esql_doc_partitioning	5590.16	ms
50th percentile service time	topnby_by_payment_type_esql_doc_partitioning	4651.93	ms
90th percentile service time	topnby_by_payment_type_esql_doc_partitioning	5207.24	ms
100th percentile service time	topnby_by_payment_type_esql_doc_partitioning	5590.16	ms
error rate	topnby_by_payment_type_esql_doc_partitioning	0	%
Min Throughput	topnby_by_payment_type_vendor_id_esql_segment_partitioning	0.14	ops/s
Mean Throughput	topnby_by_payment_type_vendor_id_esql_segment_partitioning	0.15	ops/s
Median Throughput	topnby_by_payment_type_vendor_id_esql_segment_partitioning	0.15	ops/s
Max Throughput	topnby_by_payment_type_vendor_id_esql_segment_partitioning	0.15	ops/s
50th percentile latency	topnby_by_payment_type_vendor_id_esql_segment_partitioning	6787.58	ms
90th percentile latency	topnby_by_payment_type_vendor_id_esql_segment_partitioning	7080.68	ms
100th percentile latency	topnby_by_payment_type_vendor_id_esql_segment_partitioning	7441.94	ms
50th percentile service time	topnby_by_payment_type_vendor_id_esql_segment_partitioning	6787.58	ms
90th percentile service time	topnby_by_payment_type_vendor_id_esql_segment_partitioning	7080.68	ms
100th percentile service time	topnby_by_payment_type_vendor_id_esql_segment_partitioning	7441.94	ms
error rate	topnby_by_payment_type_vendor_id_esql_segment_partitioning	0	%
Min Throughput	topnby_by_payment_type_vendor_id_esql_doc_partitioning	0.15	ops/s
Mean Throughput	topnby_by_payment_type_vendor_id_esql_doc_partitioning	0.15	ops/s
Median Throughput	topnby_by_payment_type_vendor_id_esql_doc_partitioning	0.15	ops/s
Max Throughput	topnby_by_payment_type_vendor_id_esql_doc_partitioning	0.15	ops/s
50th percentile latency	topnby_by_payment_type_vendor_id_esql_doc_partitioning	6705.6	ms
90th percentile latency	topnby_by_payment_type_vendor_id_esql_doc_partitioning	6901.74	ms
100th percentile latency	topnby_by_payment_type_vendor_id_esql_doc_partitioning	7205.46	ms
50th percentile service time	topnby_by_payment_type_vendor_id_esql_doc_partitioning	6705.6	ms
90th percentile service time	topnby_by_payment_type_vendor_id_esql_doc_partitioning	6901.74	ms
100th percentile service time	topnby_by_payment_type_vendor_id_esql_doc_partitioning	7205.46	ms
error rate	topnby_by_payment_type_vendor_id_esql_doc_partitioning	0	%

[Copilot]

Pull request overview

Adds new benchmark operations/challenges to measure ES|QL’s upcoming LIMIT ... BY behavior, including comparisons against an equivalent-ish DSL approach in the nyc_taxis track and scalability coverage in the esql track.

Changes:

• Add LIMIT ... BY ES|QL operations for nyc_taxis (segment vs doc partitioning) and corresponding challenge schedule entries.
• Add LIMIT ... BY key_1000 ES|QL operations to the esql track and wire them into the esql_large_schema challenge.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

File	Description
nyc_taxis/operations/default.json	Adds DSL + ES
nyc_taxis/challenges/default.json	Adds the new LIMIT BY operations to the nyc_taxis challenge schedule.
esql/operations/default.json	Adds ES
esql/challenges/default.json	Runs the new LIMIT ... BY operations in the esql_large_schema challenge.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

You can also share your feedback on Copilot code review. Take the survey.

[Copilot]

Using collapse with inner_hits.size: 1000 returns the top collapsed hit plus up to 1000 inner hits per payment_type (i.e., up to 1001 docs/group), which isn’t directly comparable to LIMIT 1000 BY payment_type. Also, payment_type isn’t included in inner_hits.fields, so most returned hits won’t carry the grouping key. Consider adjusting inner_hits.size (e.g., 999) and/or aligning the returned fields so this query benchmark matches the ESQL benchmark’s cardinality/projection more closely.

Suggested change

	"size": 1000,
	"sort": [{ "pickup_datetime": "desc" }],
	"_source": false,
	"fields": ["pickup_datetime", "dropoff_datetime", "trip_distance"]
	"size": 999,
	"sort": [{ "pickup_datetime": "desc" }],
	"_source": false,
	"fields": ["pickup_datetime", "dropoff_datetime", "trip_distance", "payment_type"]