rust/greptimedb
1
0
Fork 0
mirror of https://github.com/GreptimeTeam/greptimedb.git synced 2026-10-12 06:52:34 +00:00
Code Issues Packages Projects Releases Wiki Activity

test(perf): add controlled three-way join-order SQL comparison

Browse Source 
Signed-off-by: discord9 <55937128+discord9@users.noreply.github.com>
This commit is contained in:
discord9
2026-10-10 14:40:22 +08:00
parent a907cbfbe4
commit 2b3b7eab63
1 changed files with 205 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/perf/query_cases/paper_corrbound_three_way/case.toml
+205
View File
@@ -0,0 +1,205 @@
# Controlled three-way join-order comparison using the existing direct-SST DSL.
# Each table occupies one shared-layout region. Integer payloads are non-NULL;
# this case does not cover nullable join-key or payload statistics. Written join
# parentheses request SQL trees only, not optimizer hints. Timed queries return
# complete rows without aggregation or floating-point values; the runner compares
# the same query on base and candidate and reports HTTP wall-clock latency through complete response parsing, in query-block order.
# Full answer equality and final plan/operator statistics must be checked against
# an independent oracle; this fixture alone does not assert plan shape or NULLs.
[case]
name = "paper_corrbound_three_way"
description = "Three-way SQL join orders under aligned and anti-correlated filters"
[scenario]
kind = "direct_readable_sst"
seed = 20240601
[[scenario.tables]]
database = "public"
name = "paper_corrbound_r"
engine = "mito"
append_mode = true
sst_format = "flat"
primary_key = ["host"]
time_index = "ts"
[[scenario.tables.columns]]
name = "host"
type = "STRING"
semantic = "tag"
distribution = { kind = "cardinality", values = 16, prefix = "h" }
[[scenario.tables.columns]]
name = "rid"
type = "UINT64"
semantic = "field"
[[scenario.tables.columns]]
name = "ts"
type = "TIMESTAMP(9)"
semantic = "timestamp"
[[scenario.tables]]
database = "public"
name = "paper_corrbound_s"
engine = "mito"
append_mode = true
sst_format = "flat"
primary_key = ["host"]
time_index = "ts"
[[scenario.tables.columns]]
name = "host"
type = "STRING"
semantic = "tag"
distribution = { kind = "cardinality", values = 16, prefix = "h" }
[[scenario.tables.columns]]
name = "sid"
type = "UINT64"
semantic = "field"
[[scenario.tables.columns]]
name = "ts"
type = "TIMESTAMP(9)"
semantic = "timestamp"
[[scenario.tables]]
database = "public"
name = "paper_corrbound_t"
engine = "mito"
append_mode = true
sst_format = "flat"
primary_key = ["host"]
time_index = "ts"
[[scenario.tables.columns]]
name = "host"
type = "STRING"
semantic = "tag"
distribution = { kind = "cardinality", values = 2048, prefix = "h" }
[[scenario.tables.columns]]
name = "tid"
type = "UINT64"
semantic = "field"
[[scenario.tables.columns]]
name = "ts"
type = "TIMESTAMP(9)"
semantic = "timestamp"
[scenario.layout]
regions = 1
sst_count = 4
rows_per_sst = 512
row_group_size = 128
series_count = 2048
start_unix_nanos = 1704067200000000000
step_nanos = 1000000000
time_range_layout = "non_overlapping_per_sst"
series_layout = "round_robin"
[[scenario.queries]]
name = "rs_first_aligned"
kind = "sql"
query = "SELECT r.host AS r_key, r.rid AS r_val, s.host AS s_key, s.sid AS s_val, t.host AS t_key, t.tid AS t_val FROM (paper_corrbound_r AS r JOIN paper_corrbound_s AS s ON r.host = s.host) JOIN paper_corrbound_t AS t ON s.host = t.host WHERE r.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND r.ts < '2024-01-01 00:18:08+0000'::TIMESTAMP AND (r.ts < '2024-01-01 00:01:04+0000'::TIMESTAMP OR r.host = 'h0') AND s.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND s.ts < '2024-01-01 00:18:08+0000'::TIMESTAMP AND (s.ts < '2024-01-01 00:01:04+0000'::TIMESTAMP OR s.host = 'h0') AND t.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND t.ts < '2024-01-01 00:34:08+0000'::TIMESTAMP ORDER BY r_key ASC NULLS LAST, r_val ASC NULLS LAST, s_key ASC NULLS LAST, s_val ASC NULLS LAST, t_key ASC NULLS LAST, t_val ASC NULLS LAST"
warmup = 3
iterations = 9
[scenario.queries.thresholds]
max_candidate_latency_regression_pct = 50
[[scenario.queries]]
name = "t_first_aligned"
kind = "sql"
query = "SELECT r.host AS r_key, r.rid AS r_val, s.host AS s_key, s.sid AS s_val, t.host AS t_key, t.tid AS t_val FROM paper_corrbound_t AS t JOIN (paper_corrbound_s AS s JOIN paper_corrbound_r AS r ON s.host = r.host) ON s.host = t.host WHERE r.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND r.ts < '2024-01-01 00:18:08+0000'::TIMESTAMP AND (r.ts < '2024-01-01 00:01:04+0000'::TIMESTAMP OR r.host = 'h0') AND s.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND s.ts < '2024-01-01 00:18:08+0000'::TIMESTAMP AND (s.ts < '2024-01-01 00:01:04+0000'::TIMESTAMP OR s.host = 'h0') AND t.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND t.ts < '2024-01-01 00:34:08+0000'::TIMESTAMP ORDER BY r_key ASC NULLS LAST, r_val ASC NULLS LAST, s_key ASC NULLS LAST, s_val ASC NULLS LAST, t_key ASC NULLS LAST, t_val ASC NULLS LAST"
warmup = 3
iterations = 9
[scenario.queries.thresholds]
max_candidate_latency_regression_pct = 50
[[scenario.queries]]
name = "rs_first_anti"
kind = "sql"
query = "SELECT r.host AS r_key, r.rid AS r_val, s.host AS s_key, s.sid AS s_val, t.host AS t_key, t.tid AS t_val FROM (paper_corrbound_r AS r JOIN paper_corrbound_s AS s ON r.host = s.host) JOIN paper_corrbound_t AS t ON s.host = t.host WHERE r.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND r.ts < '2024-01-01 00:18:08+0000'::TIMESTAMP AND (r.ts < '2024-01-01 00:01:04+0000'::TIMESTAMP OR r.host = 'h0') AND s.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND s.ts < '2024-01-01 00:18:08+0000'::TIMESTAMP AND (s.ts < '2024-01-01 00:01:04+0000'::TIMESTAMP OR s.host = 'h1') AND t.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND t.ts < '2024-01-01 00:34:08+0000'::TIMESTAMP ORDER BY r_key ASC NULLS LAST, r_val ASC NULLS LAST, s_key ASC NULLS LAST, s_val ASC NULLS LAST, t_key ASC NULLS LAST, t_val ASC NULLS LAST"
warmup = 3
iterations = 9
[scenario.queries.thresholds]
max_candidate_latency_regression_pct = 50
[[scenario.queries]]
name = "t_first_anti"
kind = "sql"
query = "SELECT r.host AS r_key, r.rid AS r_val, s.host AS s_key, s.sid AS s_val, t.host AS t_key, t.tid AS t_val FROM paper_corrbound_t AS t JOIN (paper_corrbound_s AS s JOIN paper_corrbound_r AS r ON s.host = r.host) ON s.host = t.host WHERE r.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND r.ts < '2024-01-01 00:18:08+0000'::TIMESTAMP AND (r.ts < '2024-01-01 00:01:04+0000'::TIMESTAMP OR r.host = 'h0') AND s.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND s.ts < '2024-01-01 00:18:08+0000'::TIMESTAMP AND (s.ts < '2024-01-01 00:01:04+0000'::TIMESTAMP OR s.host = 'h1') AND t.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND t.ts < '2024-01-01 00:34:08+0000'::TIMESTAMP ORDER BY r_key ASC NULLS LAST, r_val ASC NULLS LAST, s_key ASC NULLS LAST, s_val ASC NULLS LAST, t_key ASC NULLS LAST, t_val ASC NULLS LAST"
warmup = 3
iterations = 9
[scenario.queries.thresholds]
max_candidate_latency_regression_pct = 50
[[scenario.queries]]
name = "diagnostic_explain_rs_first_aligned"
kind = "sql"
query = "EXPLAIN VERBOSE SELECT r.host AS r_key, r.rid AS r_val, s.host AS s_key, s.sid AS s_val, t.host AS t_key, t.tid AS t_val FROM (paper_corrbound_r AS r JOIN paper_corrbound_s AS s ON r.host = s.host) JOIN paper_corrbound_t AS t ON s.host = t.host WHERE r.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND r.ts < '2024-01-01 00:18:08+0000'::TIMESTAMP AND (r.ts < '2024-01-01 00:01:04+0000'::TIMESTAMP OR r.host = 'h0') AND s.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND s.ts < '2024-01-01 00:18:08+0000'::TIMESTAMP AND (s.ts < '2024-01-01 00:01:04+0000'::TIMESTAMP OR s.host = 'h0') AND t.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND t.ts < '2024-01-01 00:34:08+0000'::TIMESTAMP ORDER BY r_key ASC NULLS LAST, r_val ASC NULLS LAST, s_key ASC NULLS LAST, s_val ASC NULLS LAST, t_key ASC NULLS LAST, t_val ASC NULLS LAST"
warmup = 0
iterations = 1
[[scenario.queries]]
name = "diagnostic_explain_t_first_aligned"
kind = "sql"
query = "EXPLAIN VERBOSE SELECT r.host AS r_key, r.rid AS r_val, s.host AS s_key, s.sid AS s_val, t.host AS t_key, t.tid AS t_val FROM paper_corrbound_t AS t JOIN (paper_corrbound_s AS s JOIN paper_corrbound_r AS r ON s.host = r.host) ON s.host = t.host WHERE r.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND r.ts < '2024-01-01 00:18:08+0000'::TIMESTAMP AND (r.ts < '2024-01-01 00:01:04+0000'::TIMESTAMP OR r.host = 'h0') AND s.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND s.ts < '2024-01-01 00:18:08+0000'::TIMESTAMP AND (s.ts < '2024-01-01 00:01:04+0000'::TIMESTAMP OR s.host = 'h0') AND t.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND t.ts < '2024-01-01 00:34:08+0000'::TIMESTAMP ORDER BY r_key ASC NULLS LAST, r_val ASC NULLS LAST, s_key ASC NULLS LAST, s_val ASC NULLS LAST, t_key ASC NULLS LAST, t_val ASC NULLS LAST"
warmup = 0
iterations = 1
[[scenario.queries]]
name = "diagnostic_explain_rs_first_anti"
kind = "sql"
query = "EXPLAIN VERBOSE SELECT r.host AS r_key, r.rid AS r_val, s.host AS s_key, s.sid AS s_val, t.host AS t_key, t.tid AS t_val FROM (paper_corrbound_r AS r JOIN paper_corrbound_s AS s ON r.host = s.host) JOIN paper_corrbound_t AS t ON s.host = t.host WHERE r.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND r.ts < '2024-01-01 00:18:08+0000'::TIMESTAMP AND (r.ts < '2024-01-01 00:01:04+0000'::TIMESTAMP OR r.host = 'h0') AND s.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND s.ts < '2024-01-01 00:18:08+0000'::TIMESTAMP AND (s.ts < '2024-01-01 00:01:04+0000'::TIMESTAMP OR s.host = 'h1') AND t.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND t.ts < '2024-01-01 00:34:08+0000'::TIMESTAMP ORDER BY r_key ASC NULLS LAST, r_val ASC NULLS LAST, s_key ASC NULLS LAST, s_val ASC NULLS LAST, t_key ASC NULLS LAST, t_val ASC NULLS LAST"
warmup = 0
iterations = 1
[[scenario.queries]]
name = "diagnostic_explain_t_first_anti"
kind = "sql"
query = "EXPLAIN VERBOSE SELECT r.host AS r_key, r.rid AS r_val, s.host AS s_key, s.sid AS s_val, t.host AS t_key, t.tid AS t_val FROM paper_corrbound_t AS t JOIN (paper_corrbound_s AS s JOIN paper_corrbound_r AS r ON s.host = r.host) ON s.host = t.host WHERE r.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND r.ts < '2024-01-01 00:18:08+0000'::TIMESTAMP AND (r.ts < '2024-01-01 00:01:04+0000'::TIMESTAMP OR r.host = 'h0') AND s.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND s.ts < '2024-01-01 00:18:08+0000'::TIMESTAMP AND (s.ts < '2024-01-01 00:01:04+0000'::TIMESTAMP OR s.host = 'h1') AND t.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND t.ts < '2024-01-01 00:34:08+0000'::TIMESTAMP ORDER BY r_key ASC NULLS LAST, r_val ASC NULLS LAST, s_key ASC NULLS LAST, s_val ASC NULLS LAST, t_key ASC NULLS LAST, t_val ASC NULLS LAST"
warmup = 0
iterations = 1
[[scenario.queries]]
name = "diagnostic_analyze_rs_first_aligned"
kind = "sql"
query = "EXPLAIN ANALYZE VERBOSE SELECT r.host AS r_key, r.rid AS r_val, s.host AS s_key, s.sid AS s_val, t.host AS t_key, t.tid AS t_val FROM (paper_corrbound_r AS r JOIN paper_corrbound_s AS s ON r.host = s.host) JOIN paper_corrbound_t AS t ON s.host = t.host WHERE r.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND r.ts < '2024-01-01 00:18:08+0000'::TIMESTAMP AND (r.ts < '2024-01-01 00:01:04+0000'::TIMESTAMP OR r.host = 'h0') AND s.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND s.ts < '2024-01-01 00:18:08+0000'::TIMESTAMP AND (s.ts < '2024-01-01 00:01:04+0000'::TIMESTAMP OR s.host = 'h0') AND t.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND t.ts < '2024-01-01 00:34:08+0000'::TIMESTAMP ORDER BY r_key ASC NULLS LAST, r_val ASC NULLS LAST, s_key ASC NULLS LAST, s_val ASC NULLS LAST, t_key ASC NULLS LAST, t_val ASC NULLS LAST"
warmup = 0
iterations = 1
[[scenario.queries]]
name = "diagnostic_analyze_t_first_aligned"
kind = "sql"
query = "EXPLAIN ANALYZE VERBOSE SELECT r.host AS r_key, r.rid AS r_val, s.host AS s_key, s.sid AS s_val, t.host AS t_key, t.tid AS t_val FROM paper_corrbound_t AS t JOIN (paper_corrbound_s AS s JOIN paper_corrbound_r AS r ON s.host = r.host) ON s.host = t.host WHERE r.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND r.ts < '2024-01-01 00:18:08+0000'::TIMESTAMP AND (r.ts < '2024-01-01 00:01:04+0000'::TIMESTAMP OR r.host = 'h0') AND s.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND s.ts < '2024-01-01 00:18:08+0000'::TIMESTAMP AND (s.ts < '2024-01-01 00:01:04+0000'::TIMESTAMP OR s.host = 'h0') AND t.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND t.ts < '2024-01-01 00:34:08+0000'::TIMESTAMP ORDER BY r_key ASC NULLS LAST, r_val ASC NULLS LAST, s_key ASC NULLS LAST, s_val ASC NULLS LAST, t_key ASC NULLS LAST, t_val ASC NULLS LAST"
warmup = 0
iterations = 1
[[scenario.queries]]
name = "diagnostic_analyze_rs_first_anti"
kind = "sql"
query = "EXPLAIN ANALYZE VERBOSE SELECT r.host AS r_key, r.rid AS r_val, s.host AS s_key, s.sid AS s_val, t.host AS t_key, t.tid AS t_val FROM (paper_corrbound_r AS r JOIN paper_corrbound_s AS s ON r.host = s.host) JOIN paper_corrbound_t AS t ON s.host = t.host WHERE r.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND r.ts < '2024-01-01 00:18:08+0000'::TIMESTAMP AND (r.ts < '2024-01-01 00:01:04+0000'::TIMESTAMP OR r.host = 'h0') AND s.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND s.ts < '2024-01-01 00:18:08+0000'::TIMESTAMP AND (s.ts < '2024-01-01 00:01:04+0000'::TIMESTAMP OR s.host = 'h1') AND t.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND t.ts < '2024-01-01 00:34:08+0000'::TIMESTAMP ORDER BY r_key ASC NULLS LAST, r_val ASC NULLS LAST, s_key ASC NULLS LAST, s_val ASC NULLS LAST, t_key ASC NULLS LAST, t_val ASC NULLS LAST"
warmup = 0
iterations = 1
[[scenario.queries]]
name = "diagnostic_analyze_t_first_anti"
kind = "sql"
query = "EXPLAIN ANALYZE VERBOSE SELECT r.host AS r_key, r.rid AS r_val, s.host AS s_key, s.sid AS s_val, t.host AS t_key, t.tid AS t_val FROM paper_corrbound_t AS t JOIN (paper_corrbound_s AS s JOIN paper_corrbound_r AS r ON s.host = r.host) ON s.host = t.host WHERE r.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND r.ts < '2024-01-01 00:18:08+0000'::TIMESTAMP AND (r.ts < '2024-01-01 00:01:04+0000'::TIMESTAMP OR r.host = 'h0') AND s.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND s.ts < '2024-01-01 00:18:08+0000'::TIMESTAMP AND (s.ts < '2024-01-01 00:01:04+0000'::TIMESTAMP OR s.host = 'h1') AND t.ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND t.ts < '2024-01-01 00:34:08+0000'::TIMESTAMP ORDER BY r_key ASC NULLS LAST, r_val ASC NULLS LAST, s_key ASC NULLS LAST, s_val ASC NULLS LAST, t_key ASC NULLS LAST, t_val ASC NULLS LAST"
warmup = 0
iterations = 1
[[scenario.queries]]
name = "diagnostic_source_counts"
kind = "sql"
query = "SELECT 'R' AS source, COUNT(*) AS row_count FROM paper_corrbound_r WHERE ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND ts < '2024-01-01 00:18:08+0000'::TIMESTAMP AND (ts < '2024-01-01 00:01:04+0000'::TIMESTAMP OR host = 'h0') UNION ALL SELECT 'S_aligned', COUNT(*) FROM paper_corrbound_s WHERE ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND ts < '2024-01-01 00:18:08+0000'::TIMESTAMP AND (ts < '2024-01-01 00:01:04+0000'::TIMESTAMP OR host = 'h0') UNION ALL SELECT 'S_anti', COUNT(*) FROM paper_corrbound_s WHERE ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND ts < '2024-01-01 00:18:08+0000'::TIMESTAMP AND (ts < '2024-01-01 00:01:04+0000'::TIMESTAMP OR host = 'h1') UNION ALL SELECT 'T', COUNT(*) FROM paper_corrbound_t WHERE ts >= '2024-01-01 00:00:00+0000'::TIMESTAMP AND ts < '2024-01-01 00:34:08+0000'::TIMESTAMP ORDER BY source"
warmup = 0
iterations = 1