bench: add the Star Schema Benchmark to the SQL bench matrix

.github/workflows/sql-bench-matrix.yml

@@ -136,6 +136,12 @@ jobs:
           chmod +x duckdb
           echo "$PWD" >> "$GITHUB_PATH"
 
+      # SSB has no Rust generator; `vx-bench prepare-data ssb` builds the reference C `dbgen`
+      # from source. See vortex-bench/sql/ssb/README.md.
+      - name: Install cmake
+        if: matrix.subcommand == 'ssb'
+        run: sudo apt-get update && sudo apt-get install -y cmake
+
       - uses: ./.github/actions/system-info
 
       - name: Download binaries

bench-orchestrator/README.md

@@ -41,7 +41,7 @@ vx-bench run <benchmark> [options]
 
 **Arguments:**
 
-- `benchmark`: Benchmark suite to run (`appian`, `tpch`, `tpcds`, `clickbench`, `fineweb`, `gh-archive`, `polarsignals`, `public-bi`, `statpopgen`)
+- `benchmark`: Benchmark suite to run (`appian`, `tpch`, `tpcds`, `clickbench`, `fineweb`, `gh-archive`, `polarsignals`, `public-bi`, `ssb`, `statpopgen`)
 
 **Options:**
 

bench-orchestrator/bench_orchestrator/ci_matrix/catalog.py

@@ -263,6 +263,19 @@
             "develop": STANDARD_WITH_DUCKDB,
         },
     ),
+    BenchmarkCase(
+        id="ssb-nvme",
+        benchmark=Benchmark.SSB,
+        name="SSB SF=10 on NVME",
+        scale_factor=10.0,
+        iterations=10,
+        runs={
+            "pr-compact": COMPACT,
+            "pr-all": COMPACT,
+            "pr-full": DEFAULT_WITH_DUCKDB_PR_FULL,
+            "develop": STANDARD_WITH_DUCKDB,
+        },
+    ),
     BenchmarkCase(
         id="vortex-queries",
         benchmark=Benchmark.VORTEX_QUERIES,

bench-orchestrator/bench_orchestrator/config.py

@@ -62,6 +62,7 @@ class Benchmark(Enum):
     PUBLIC_BI = "public-bi"
     STATPOPGEN = "statpopgen"
     SPATIALBENCH = "spatialbench"
+    SSB = "ssb"
     VORTEX_QUERIES = "vortex"
 
 

bench-orchestrator/tests/test_matrix.py

@@ -31,6 +31,7 @@
     "fineweb-s3",
     "polarsignals",
     "appian-nvme",
+    "ssb-nvme",
     "vortex-queries",
 )
 COMPACT_IDS = tuple(
@@ -42,7 +43,7 @@
     "pr": tuple(
         benchmark_id
         for benchmark_id in REGULAR_IDS
-        if benchmark_id not in {"tpch-s3-10", "appian-nvme", "vortex-queries"}
+        if benchmark_id not in {"tpch-s3-10", "appian-nvme", "ssb-nvme", "vortex-queries"}
     ),
     "pr-compact": COMPACT_IDS,
     "pr-all": PR_ALL_IDS,

vortex-bench/sql/ssb/README.md

@@ -0,0 +1,67 @@
+# Star Schema Benchmark
+
+The [Star Schema Benchmark](https://www.cs.umb.edu/~poneil/StarSchemaB.PDF) (O'Neil, O'Neil &
+Chen) is TPC-H redesigned as a textbook star schema: TPC-H's `lineitem` and `orders` are
+denormalized into one wide `lineorder` fact table, joined against four dimensions — `customer`,
+`supplier`, `part`, and `dwdate`.
+
+That shape is what makes it worth running alongside TPC-H. Every query is a scan of the fact
+table under dimension-derived filters of a known, deliberately varied selectivity, so the suite
+isolates filter pushdown, zone-map pruning, and dimension-join throughput rather than mixing them
+with TPC-H's subqueries and correlated predicates.
+
+## Queries
+
+The 13 queries are organized into four "flights", each holding the query shape fixed while
+tightening the filters. The harness keys queries on a plain index, so the paper's numbering maps
+onto `q1.sql` ... `q13.sql` in order:
+
+| File | SSB | Flight |
+| --- | --- | --- |
+| `q1.sql` | Q1.1 | Flight 1 — single-dimension discount/quantity filter on `lineorder` x `dwdate` |
+| `q2.sql` | Q1.2 | |
+| `q3.sql` | Q1.3 | |
+| `q4.sql` | Q2.1 | Flight 2 — `part` and `supplier` restriction, grouped by year and brand |
+| `q5.sql` | Q2.2 | |
+| `q6.sql` | Q2.3 | |
+| `q7.sql` | Q3.1 | Flight 3 — `customer` x `supplier` geography join, narrowing region to nation to city |
+| `q8.sql` | Q3.2 | |
+| `q9.sql` | Q3.3 | |
+| `q10.sql` | Q3.4 | |
+| `q11.sql` | Q4.1 | Flight 4 — all four dimensions, profit aggregation |
+| `q12.sql` | Q4.2 | |
+| `q13.sql` | Q4.3 | |
+
+Each file names its SSB query in a leading comment.
+
+## Schema notes
+
+The date dimension is registered as **`dwdate`**, not `date`: `date` is a reserved word in both
+DataFusion's and DuckDB's parsers. This is the same rename the reference SSB load scripts apply,
+for the same reason. Column names (`d_datekey`, `d_year`, ...) are unchanged.
+
+## Data
+
+There is no Rust SSB generator, and SSB is not derivable from TPC-H output — the dimension
+cardinalities differ (`customer` is SF x 30k rather than SF x 150k, `supplier` SF x 2k rather than
+SF x 10k, `part` is `200000 * floor(1 + log2(SF))`) and `dwdate` has no TPC-H analogue. So
+[`src/ssb/datagen.rs`](../../src/ssb/datagen.rs) clones and builds the pinned reference C `dbgen`,
+runs it, and converts its `.tbl` output to Parquet with the `duckdb` CLI. Generating data
+therefore needs a C compiler, `cmake`, `git`, and the `duckdb` CLI on `PATH`; everything is
+cached and idempotent after the first run.
+
+SSB has no official upstream, only a tree of unsynchronized `dbgen` forks, and they are not
+interchangeable — the module docs record which fork is pinned and why. At SF 10 the source
+Parquet is ~1.7 GB, dominated by `lineorder`'s 59,986,217 rows.
+
+## CI variant
+
+CI runs this suite at scale factor 10 from local NVMe as the `SSB SF=10 on NVME` PR comment,
+comparing DataFusion and DuckDB over Parquet, Vortex, and vortex-compact files (plus a native
+DuckDB baseline). Like Appian, it is part of the full SQL matrix rather than the quick PR one.
+
+## Running locally
+
+```bash
+vx-bench run ssb --engine datafusion,duckdb --format parquet,vortex --opt scale-factor=10.0
+```

vortex-bench/sql/ssb/q1.sql

@@ -0,0 +1,11 @@
+-- SSB Q1.1
+select
+    sum(lo_extendedprice * lo_discount) as revenue
+from
+    lineorder,
+    dwdate
+where
+        lo_orderdate = d_datekey
+    and d_year = 1993
+    and lo_discount between 1 and 3
+    and lo_quantity < 25

vortex-bench/sql/ssb/q10.sql

@@ -0,0 +1,25 @@
+-- SSB Q3.4
+select
+    c_city,
+    s_city,
+    d_year,
+    sum(lo_revenue) as lo_revenue
+from
+    customer,
+    lineorder,
+    supplier,
+    dwdate
+where
+        lo_custkey = c_custkey
+    and lo_suppkey = s_suppkey
+    and lo_orderdate = d_datekey
+    and (c_city = 'UNITED KI1' or c_city = 'UNITED KI5')
+    and (s_city = 'UNITED KI1' or s_city = 'UNITED KI5')
+    and d_yearmonth = 'Dec1997'
+group by
+    c_city,
+    s_city,
+    d_year
+order by
+    d_year asc,
+    lo_revenue desc

vortex-bench/sql/ssb/q11.sql

@@ -0,0 +1,25 @@
+-- SSB Q4.1
+select
+    d_year,
+    c_nation,
+    sum(lo_revenue - lo_supplycost) as profit
+from
+    dwdate,
+    customer,
+    supplier,
+    part,
+    lineorder
+where
+        lo_custkey = c_custkey
+    and lo_suppkey = s_suppkey
+    and lo_partkey = p_partkey
+    and lo_orderdate = d_datekey
+    and c_region = 'AMERICA'
+    and s_region = 'AMERICA'
+    and (p_mfgr = 'MFGR#1' or p_mfgr = 'MFGR#2')
+group by
+    d_year,
+    c_nation
+order by
+    d_year,
+    c_nation

vortex-bench/sql/ssb/q12.sql

@@ -0,0 +1,29 @@
+-- SSB Q4.2
+select
+    d_year,
+    s_nation,
+    p_category,
+    sum(lo_revenue - lo_supplycost) as profit
+from
+    dwdate,
+    customer,
+    supplier,
+    part,
+    lineorder
+where
+        lo_custkey = c_custkey
+    and lo_suppkey = s_suppkey
+    and lo_partkey = p_partkey
+    and lo_orderdate = d_datekey
+    and c_region = 'AMERICA'
+    and s_region = 'AMERICA'
+    and (d_year = 1997 or d_year = 1998)
+    and (p_mfgr = 'MFGR#1' or p_mfgr = 'MFGR#2')
+group by
+    d_year,
+    s_nation,
+    p_category
+order by
+    d_year,
+    s_nation,
+    p_category

vortex-bench/sql/ssb/q13.sql

@@ -0,0 +1,28 @@
+-- SSB Q4.3
+select
+    d_year,
+    s_city,
+    p_brand1,
+    sum(lo_revenue - lo_supplycost) as profit
+from
+    dwdate,
+    customer,
+    supplier,
+    part,
+    lineorder
+where
+        lo_custkey = c_custkey
+    and lo_suppkey = s_suppkey
+    and lo_partkey = p_partkey
+    and lo_orderdate = d_datekey
+    and s_nation = 'UNITED STATES'
+    and (d_year = 1997 or d_year = 1998)
+    and p_category = 'MFGR#14'
+group by
+    d_year,
+    s_city,
+    p_brand1
+order by
+    d_year,
+    s_city,
+    p_brand1

vortex-bench/sql/ssb/q2.sql

@@ -0,0 +1,11 @@
+-- SSB Q1.2
+select
+    sum(lo_extendedprice * lo_discount) as revenue
+from
+    lineorder,
+    dwdate
+where
+        lo_orderdate = d_datekey
+    and d_yearmonthnum = 199401
+    and lo_discount between 4 and 6
+    and lo_quantity between 26 and 35

vortex-bench/sql/ssb/q3.sql

@@ -0,0 +1,12 @@
+-- SSB Q1.3
+select
+    sum(lo_extendedprice * lo_discount) as revenue
+from
+    lineorder,
+    dwdate
+where
+        lo_orderdate = d_datekey
+    and d_weeknuminyear = 6
+    and d_year = 1994
+    and lo_discount between 5 and 7
+    and lo_quantity between 26 and 35

vortex-bench/sql/ssb/q4.sql

@@ -0,0 +1,22 @@
+-- SSB Q2.1
+select
+    sum(lo_revenue) as lo_revenue,
+    d_year,
+    p_brand1
+from
+    lineorder,
+    dwdate,
+    part,
+    supplier
+where
+        lo_orderdate = d_datekey
+    and lo_partkey = p_partkey
+    and lo_suppkey = s_suppkey
+    and p_category = 'MFGR#12'
+    and s_region = 'AMERICA'
+group by
+    d_year,
+    p_brand1
+order by
+    d_year,
+    p_brand1

vortex-bench/sql/ssb/q5.sql

@@ -0,0 +1,22 @@
+-- SSB Q2.2
+select
+    sum(lo_revenue) as lo_revenue,
+    d_year,
+    p_brand1
+from
+    lineorder,
+    dwdate,
+    part,
+    supplier
+where
+        lo_orderdate = d_datekey
+    and lo_partkey = p_partkey
+    and lo_suppkey = s_suppkey
+    and p_brand1 between 'MFGR#2221' and 'MFGR#2228'
+    and s_region = 'ASIA'
+group by
+    d_year,
+    p_brand1
+order by
+    d_year,
+    p_brand1

vortex-bench/sql/ssb/q6.sql

@@ -0,0 +1,22 @@
+-- SSB Q2.3
+select
+    sum(lo_revenue) as lo_revenue,
+    d_year,
+    p_brand1
+from
+    lineorder,
+    dwdate,
+    part,
+    supplier
+where
+        lo_orderdate = d_datekey
+    and lo_partkey = p_partkey
+    and lo_suppkey = s_suppkey
+    and p_brand1 = 'MFGR#2239'
+    and s_region = 'EUROPE'
+group by
+    d_year,
+    p_brand1
+order by
+    d_year,
+    p_brand1

vortex-bench/sql/ssb/q7.sql

@@ -0,0 +1,26 @@
+-- SSB Q3.1
+select
+    c_nation,
+    s_nation,
+    d_year,
+    sum(lo_revenue) as lo_revenue
+from
+    customer,
+    lineorder,
+    supplier,
+    dwdate
+where
+        lo_custkey = c_custkey
+    and lo_suppkey = s_suppkey
+    and lo_orderdate = d_datekey
+    and c_region = 'ASIA'
+    and s_region = 'ASIA'
+    and d_year >= 1992
+    and d_year <= 1997
+group by
+    c_nation,
+    s_nation,
+    d_year
+order by
+    d_year asc,
+    lo_revenue desc