Illustrative sample. Representative of a completed EWIS AEGIS engagement
A legacy IT Help Desk application with per-user Oracle-credential login, an in-memory real-time job board, and stringly-typed data contracts was rebuilt as a modern three-tier platform. The proven Oracle business logic, which encodes years of operational knowledge, was deliberately retained: "retain the proven Oracle logic, replace everything around it."
At a Glance
| Metric | Result |
|---|---|
| Largest Contentful Paint | 1.2s (from ~3.2s legacy) |
| API latency (p95) | <200ms, pooled & typed access |
| Concurrent real-time connections | ~100k per SignalR instance |
| Identity attack reduction | 99.2% fewer (Entra ID + MFA) |
Key Transformations
- Three-tier architecture, Oracle logic untouched: A Next.js front end and dedicated Node.js API tier (BFF) sit in front of the existing Oracle package through an anti-corruption layer. The database never learns about HTTP; the front end never touches Oracle directly.
- Identity modernization: Microsoft Entra ID (OIDC + MFA) replaced personal Oracle credentials for login. That closed a SQL-injectable login path and removed database credential handling from the app entirely.
- Real-time delivery at scale: The legacy in-memory, single-server job board was replaced with Azure SignalR Service, a managed backplane that removes the sticky-session constraint and scales horizontally.
- Typed data contracts: Legacy positional string arrays (
List<List<String>>) were replaced with strongly-typed TypeScript domain models, assembled by the anti-corruption layer from the retained Oracle procedures. - Pooled database access: A one-session-per-user Oracle connection model was replaced with a fixed-size connection pool (per Oracle's Real-World Performance guidance), with per-request user identity stamped for accurate database-side auditing.
- Three new business modules delivered: Problem Management, a Task Breakdown Tracker, and a Reporting module (on a dedicated, event-fed read model) were added alongside a redesigned, SLA-centric real-time dashboard.
- Zero-downtime rollout: Delivered behind a strangler-fig façade: the legacy system kept running until each replacement capability was proven in production, so cutover was incremental and reversible instead of a single high-risk switch.
Key Benefits
| Area | Before | After |
|---|---|---|
| First paint (LCP) | ~3.2s | ~1.2s |
| API latency (p95) | often >500ms | <200ms |
| Real-time scale | A few thousand, single server | ~100k per instance, multi-instance |
| Authentication | Personal Oracle credentials | Entra ID SSO + MFA |
| Test coverage | ~0% | ~80% (ratcheted on new code) |
| Deployment | Infrequent, manual | On-demand CI/CD |
Why It Matters
Modernization doesn't require rewriting or risking the business logic that already works. By isolating the proven Oracle package behind an anti-corruption layer and modernizing identity, real-time delivery, and the presentation tier around it, the team delivered measurable gains in speed, scale, and security while keeping years of encoded operational knowledge intact.
Illustrative sample. Client-identifying details removed. Figures are representative of a comparable completed engagement, grounded in public benchmarks (Google Core Web Vitals, DORA 2024 State of DevOps, Microsoft Learn, Oracle documentation) and are not guarantees. Prepared by EWIS AEGIS.