Oninit® Informix Analyser — For Architects
The Analyser is delivered as a managed diagnostic service. There is no agent on the customer host, no rules engine for the customer to operate, and no schema change anywhere in the customer estate. The customer uploads diagnostic data to the portal; a Python application drives a SWI-Prolog rules engine over those facts and produces a report.
Architectural shape
Python orchestrates I/O and report rendering through the Janus bidirectional bridge to SWI-Prolog. The fact base is rebuilt per collection — assertz / retract cycles ensure no carry-over between runs — and the rule files are plain .pl source, reviewable and hand-editable. Findings are first-class finding/5 Prolog terms; the meta-interpreter exposes the derivation chain so every finding has a traceable evidence path.
Per-engagement state lives in a portal database. Submission directories are addressed by ULID so they sort chronologically without a separate index. Cross-tenant access is structurally prevented at the storage layer (cross-tenant reads return HTTP 404 rather than 403, so a stranger can't even probe submission existence).
Tier model
| Tier | Uploads | Monthly processing | Capabilities |
|---|---|---|---|
| Free | 100 MB cap | — | High-level analysis. Lowest queue priority. No persistence. No custom modules. |
| Basic | 500 MB | ~10 GB | Moderate analysis depth. Persistent saved reports. Public module catalogue. |
| Pro | 2 GB direct + external integration for larger files | ~50 GB | Full depth. Persistent context. High queue priority. Custom modules. Defined private-module allocation. |
| Enterprise | 200 GB+ custom quotas, pull-based integrations | Continuous ingestion | Dedicated capacity. SLA-backed performance. Unlimited private modules. RBAC. Bypasses the shared queue. |
Persistent context across sessions (Basic and above) means a configuration drift flagged in May is still visible when the same instance is analysed in November. Recurring patterns across an estate emerge over time; one-off findings don't get lost between investigations.
Integration model
The portal accepts direct upload up to 2 GB and gates larger files behind external integrations — Dropbox, Google Drive, FTP, API. Integrations are OAuth-based; no third-party credentials are stored on Oninit infrastructure. The OAuth flow returns a short-lived access token that the portal uses to fetch the upload, then discards.
Result and metadata storage is the persistent shape. Raw inputs are not retained as primary storage — once the report renders, the raw upload is purgeable per the engagement's retention policy.
Pre-analysis validation
Before a submission lands in the analysis queue, the portal flags optimisation opportunities in the collection itself — for example, recommending DUMPSHM be disabled before submission to keep file sizes proportionate to their analytical value. The collection cost is part of the architecture, not an afterthought.
Output channels
- Branded PDF report for stakeholders, with optional methodology appendix explaining how each rule derived its conclusion.
- Plain CLI report from the same renderer for pipelines and headless use.
- Structured findings exposable for downstream consumers (the same finding/5 data the renderer reads, in machine-readable form).
Module rules are reviewable Prolog source. The methodology appendix in the PDF makes derivation auditable.
Where it fits in a target architecture
- A managed diagnostic SaaS sitting alongside an existing Informix estate — no agent, no rule engine to deploy in-house.
- A long-term intelligence layer over a portfolio of Informix instances, with per-instance and cross-instance pattern recognition.
- A pre-flight / post-deploy gate during change windows — submit a collection before the change, again after, see what moved.
- A second opinion path for in-house DBA teams — the rules engine encodes Oninit's two decades of field experience as version-controlled, reviewable Prolog source.