Products › Data Quality › Metadata Scanner – Mainframe
PDI Metadata Scanner – Mainframe

Your mainframe, mapped. Every fact, proven.

A metadata catalog and column-level lineage explorer for the six classic mainframe estates. Every object and every edge carries the source line that proves it — and when the scanner can't prove something, it says Unverified instead of guessing.

100M
Lines of source metadata
scanned in 18.5 seconds
6
Mainframe estates
in one catalog
100%
Of facts source-traced
or flagged Unverified
PDI Metadata Scanner — Catalog Explorer
DB2 z/OS
DB2 for i
IMS
Adabas
CA IDMS
CA Datacom
BANKAPP.CUSTOMER — table
✓ Verified 5 columns
⌖ BANKAPP.sql:2 — CREATE TABLE BANKAPP.CUSTOMER
Where used
CUSTUPDT · EXEC SQL UPDATE · CUSTUPDT.cbl:17
BANKAPP.LEDGER
◮ Unverified — needs manual confirmation

Engineered to be believed.

A catalog is only useful if your stewards can trust it. Every capability below exists to make each fact either provable — or honestly flagged.

The Provenance Ledger

Every object and lineage edge records the source member, line number, and statement that proves it. Hover any fact and see its evidence. There is no code path that creates a fact without one.

provenance tuple required on every record

Refuses to Guess

A table with no DDL, a DL/I call resolved at runtime, a program missing from source control — these stay visible as dashed, amber-flagged edges labeled "Unverified." Silent omission and confident fabrication are both failures. The scanner commits neither.

tests assert every unprovable case surfaces
🗄️

Six Estates, One Catalog

Db2 for z/OS, Db2 for i, IMS, Adabas, CA IDMS, and CA Datacom/DB — parsed from each platform's own extraction output (catalog unloads, DBD source, FDT reports, dictionary punches), never live catalog cursors.

lineage from COBOL · JCL · Natural · IDMS DML
♻️

Zero-Drift Re-runs

Deterministic path identifiers, converge-by-key writes, and a corpus content hash. Re-scan unchanged sources and the catalog is byte-identical — proven by diff in the regression suite, not promised in a slide.

tests/test_idempotency.sh — zero-byte diff
🔐

Single Audited Secret Path

Passwords encrypted at rest (Fernet: AES-128-CBC + HMAC-SHA256), key held outside the repository. Never in the config file, the logs, the process list, or the UI after entry — the engine receives credentials over an in-memory pipe.

tests/test_secrets.py — no-plaintext invariant
📤

Informatica CDGC Export

Objects and lineage export in an Oracle-facade convention ready for CDGC bulk ingestion — with the provenance columns included, so the evidence travels with the metadata into your governance platform.

one-click CSV · headers reconciled at onboarding

Four moves. Zero ambiguity.

One decision per screen, plain-English captions on every field, and a Save button that stays disabled until the connection test passes.

1

Pick a Source

Six estate cards. Nothing else on screen. Already-connected sources carry a badge; picking one is the only decision this screen asks of you.

2

Test, Then Save

Every field explains itself ("ask your mainframe admin if unsure"). Save & Connect only unlocks after the test passes. Passwords are encrypted the moment you save.

3

The Engine Scans

The mainframe's own unload utilities extract structure on-platform behind a busy-host guardrail; the C++ engine parses source and builds the provenance-backed catalog off-host.

4

Explore & Export

Browse database → schema → table → column, open Where Used, follow the lineage flow, and export the catalog to Informatica CDGC — evidence included.

Measured. Not marketed.

Every number below is produced by the benchmark harness and test suite that ship with the product — reproducible with a single command, hardware stated in full in the whitepaper.

5.4M
lines/sec
Production scan loop
(single CPU core)
18.5s
for 100M lines
16× the 5-minute requirement
(measured, not extrapolated)
<1s
every view
The UI only reads indexed
artifacts — at any catalog size
0
bytes drift
Re-run on unchanged sources:
byte-identical catalog
"A catalog earns trust the way an audit does: not by claiming to be right, but by showing its evidence for every line."
The design principle behind the Provenance Ledger
From the PDI engineering whitepaper · customer references available on request

Go deeper.

The engineering whitepaper: the provenance ledger, the on-host/off-host offload design, and the full measured benchmark methodology.

Read it on the web

Same content, browsable on any device.

Open

Download the PDF

Print-ready, for the steering committee.

Download

See your own estate mapped.

Bring a copybook, a DBD, or a schema punch to a 30-minute briefing. We'll show you the catalog, the lineage, and — just as importantly — what it refuses to claim.

Request a Demo