Workflow Progress Storage Benchmark¶

Raw JSON is authoritative. This summary uses medians across all modeled cases and warm samples.

Environment: Windows-11-10.0.26200-SP0; Python 3.12.12; postgresql 170010.

Collected: 2026-07-20T23:42:21.378055+00:00; database deployment: docker-desktop:postgres:17@sha256:a426e44bac0b759c95894d68e1a0ac03ecc20b619f498a91aae373bf06d8508d; psycopg 3.3.4.

Source revision: unavailable; benchmark implementation SHA-256: da4616a613cf8643a0822bba451bbb8c1b8feb8d59a5048871dc9b8acff787fd.

Matrix: nodes [1000, 10000, 50000, 100000]; change rates [0.01, 0.1, 1.0]; repetitions 5; seed 20260720.

Execution scope: configured Django database only; this command does not start Ray or access Kubernetes.

Python timings cover bounded in-process structures. Write bytes and database statement counts are analytical; the representative JSONB probe does not measure candidate-schema round trips or throughput.

Candidate	Median total write bytes	Median estimated DB statements	Median warm serialize (ms)
`current_full_row`	24042048	1.0	0.5183
`bounded_inline`	16384	1.0	0.1953
`chunked_database`	10487808	112.5	0.3684
`normalized`	2320048	4.0	0.3664
`append_delta`	2230048	6.0	0.3690
`external_chunk`	10492672	2.0	0.3677
`live_only`	2048	1.0	0.0080

Sparse short-profile focus (100000 nodes, 1% changed)¶

Candidate	Modeled unit kind	Touched units	Estimated DB statements	Total write bytes
`current_full_row`	`full_graph_page_equivalent`	391	1	22502048
`bounded_inline`	`none`	0	1	6176
`chunked_database`	`expected_random_page`	361	363	16359680
`normalized`	`none`	0	4	227048
`append_delta`	`delta_chunk`	4	6	195048
`external_chunk`	`expected_random_page`	361	2	16395360
`live_only`	`none`	0	1	2048

Database evidence: available (postgresql). Representative relation growth: 49152 bytes; WAL: 10544 bytes; table cleaned: True.

Timings are observations, not CI thresholds or service-level objectives.