Legacy OSS Data Extraction Service

Problem Context

Telecom engineers maintain 20-year-old OSS systems with C++ cores, Perl glue, and MySQL databases. They need clean data for ML but can't modify the legacy codebase or poll the database during peak hours. Current extraction methods fail within weeks due to format drift and performance issues.

Pain Points

Log parsing breaks after software updates. DB polling crashes systems during high traffic. Instrumenting C++ binaries gets blocked by security teams. Manual normalization takes months to maintain. Every failed attempt wastes 50+ hours of engineering time per project.

Impact

Downtime costs $10k+/hour. ML projects get delayed 6-12 months. Engineers spend 20+ hours/week fighting data extraction instead of building features. Legacy systems become technical debt black holes that block digital transformation.

Urgency

These systems can't be replaced overnight. Every day without clean data extraction means lost revenue from stalled ML initiatives. Security teams will never approve code changes. The only path forward is non-invasive extraction that works with existing infrastructure.

Target Audience

Telecom OSS engineers, legacy system architects, enterprise IT teams maintaining 10+ year old mission-critical systems, data scientists working with telecom/finance legacy data, DevOps teams supporting non-cloud-native applications

Solution Approach

A self-deployable service that extracts clean data from legacy OSS stacks without modifying the application code. Uses Change Data Capture (CDC) on MySQL binlogs to get real-time event streams, then automatically normalizes 15+ years of format drift into a stable API for ML teams. Deploys via Docker container with no admin rights required.

Key Features

1. Automatic Normalizer: Handles 15+ years of undocumented format changes, timezone migrations, and repurposed columns.
2. Stable API Layer: Provides a consistent interface for ML teams regardless of underlying data chaos.
3. Docker Deployment: One-command setup that runs alongside existing systems without requiring admin access or code changes.

User Experience

Engineers install via Docker in 10 minutes. The service starts capturing data immediately. Normalization rules adapt automatically to new format drifts. ML teams get clean data via API without knowing the legacy system exists. No more manual ETL scripts or performance-killing DB polls.

Differentiation

Unlike generic ETL tools, this focuses specifically on legacy OSS stacks with 15+ years of undocumented format drift. The automatic normalization handles cases where columns were repurposed silently over decades. Docker deployment avoids the 'no admin access' roadblock that kills other solutions. No code changes or security sign-offs required.

Scalability

Starts with a single Docker container. Adds more containers for higher throughput. Normalization rules can be shared across similar legacy systems. API can be extended to support additional data sources as needs grow. Pricing scales with data volume and team size.

Expected Impact

ML projects ship on time instead of being delayed 6-12 months. Engineers spend 0 hours maintaining data extraction pipelines. Systems stay stable during peak loads. Legacy data becomes a competitive advantage instead of a technical debt nightmare. Teams can finally modernize without replacing working systems.