★ LANGUAGE primary ★ DEVICE I/O cross-listed

SlimeTree-VSAM — Bench procedure and primary materials

For evaluators and procurement. The measurement conditions and dataset scale behind sequential cursor 480×, random key 267×, and the 1-billion-record nightly batch 19.5h → 4.4 min measured in same-host bench vs PostgreSQL 16; the SHA-256 audit-chain regulatory alignment; and the drop-in integration procedure with SlimeCOBOL Phase D emit.

For product overview and deployment scenarios, see the product page (/products/language/slimetree-vsam/). This page focuses on primary materials for reproduction and verification.

1. Same-host bench results Open

Rather than relying on third-party benchmarks, we use same-host bench that is reproducible in the customer's environment. PostgreSQL 16 runs directly under the same conditions, with realistic operational workloads for comparison.

Metricbaseline (PostgreSQL 16)SlimeTree-VSAMRatio
per-access (sequential cursor) 1.0× (baseline) 480× faster 480×
per-access (random key) 1.0× 267× faster 267×
1B-record nightly batch (representative workload) 19.5 hours 4.4 minutes 266× shorter

1.1 Reproduction conditions

baselinePostgreSQL 16 (same host, default config + standard tuning). Version / OS / kernel / file system conditions can be aligned per contract
SlimeTree-VSAMSingle Rust binary (static link), same OS / same FS. Implements VSAM (KSDS / ESDS / RRDS) compatible access methods
workload1-billion-record-equivalent operational records (representative banking-core batch), mixed sequential-cursor full-scan + random-key retrieval
hostSame-host bench (eliminates network effects). In production: SAN / NAS / direct-attached storage all supported
metricsper-access latency / total throughput / wall clock. Concurrent jobs, cache conditions, and disk-I/O conditions explicitly stated
reproduction kitAfter NDA: bench script + sample-dataset spec + expected-value ranges provided
Stance: The point is not "480× in a public benchmark" but "480× reproducible in your environment with the same procedure". A same-condition PoC can be set up in 3-5 business days after NDA.

2. Drop-in integration with SlimeCOBOL Phase D emit Open

SlimeTree-VSAM is designed as the DB target for SlimeCOBOL Phase D (QSAM / VSAM emit). Verification confirms that the VSAM access output by Phase D is received unchanged on the SlimeTree-VSAM side.

RECFM compatibilityFixed (F) / Variable (V) / with-control (M) / Blocked (B) reproduced bit-exact
Access methodKSDS (key-sequenced) / ESDS (entry-sequenced) / RRDS (relative-record) all auto-routed by the emitter. Drop-in
EncodingEBCDIC / packed decimal / zoned decimal mainframe types preserved (no conversion)
Audit linkageSlimeCOBOL's bit-exact transcription result connects directly to SlimeTree-VSAM's SHA-256 audit chain. Transcription → storage → audit form a three-stage tamper-detection chain

3. SHA-256 audit chain Open

Every write operation on SlimeTree-VSAM is recorded as a SHA-256 chain; tamper detection is built into the backend. After-the-fact audit works in air-gap environments (no network access).

  • The Write-Ahead Log is SHA-256 chained, each record embedding the previous hash
  • Replay is deterministic — the same inputs regenerate the same chain; a divergence isolates the tamper point
  • Air-gap audit: completes inside the internal audit chain without external network
  • Aligns with bit-exact audit requirements of regulated industries (banking, insurance, public sector, healthcare)

4. Patent Coverage public only — text NDA-gated

SlimeTree-VSAM implements VSAM-standard-compatible access methods in native Rust, and the implementation-level simultaneous achievement of performance and audit is related to the SlimeRESCUE / SlimeNENC patent family. Text access via contact → after NDA.

5. Implementation (code) Distribution preparing

Rust implementationSingle Rust binary (static link). Dependencies: standard library + cryptographic hash. WASM build in preparation (restricted access-method subset, for browser evaluation)
Access methodsKSDS / ESDS / RRDS all implemented, bit-exact RECFM supported
Bench harnessSame-condition replay scripts for PostgreSQL 16 comparison, dataset generator, measurement logger, result aggregator
SlimeCOBOL integrationDrop-in connection test suite with Phase D emit, RECFM bit-exact verification

Distribution forms (evaluation licence / joint PoC / sponsored development / OEM integration) via contact or the partners page.