At a Glance
A leading Japanese mobile network provider — one of the largest telecommunications companies in Asia — faced an immovable datacenter exit deadline. With 2,500 virtual machines hosting over 1,300 business-critical applications, the internal team estimated the assessment alone would take 18 months. They had three.
Using CHAI, CloudHedge's agentic AI platform, the provider completed full discovery, assessment, and modernization planning for their entire portfolio in the required timeframe — and achieved a 50% reduction in infrastructure costs after migration.
The Situation
The provider operated a sprawling datacenter environment built up over more than a decade. Applications ranged from customer-facing mobile platforms to internal billing systems, network management tools, and legacy middleware stacks. The technology footprint included Java, .NET, Python, and C++ applications running on a mix of Windows and Linux VMs.
A corporate mandate required the complete exit of two physical datacenters within 12 months. The first phase — full discovery and assessment of every application, with a disposition recommendation and migration plan — needed to be complete within 3 months.
The internal team evaluated multiple consulting firms. Every estimate came back the same: 12 to 18 months minimum for the assessment phase alone, with teams of 30 to 50 consultants.
The Real Problem
The challenge was not just scale — it was complexity and interdependence. Many applications had undocumented dependencies. Services communicated through shared databases, message queues, and file system mounts that no architecture diagram captured. Configuration management data was incomplete. Tribal knowledge lived in the heads of engineers who had moved on.
A manual assessment would have required consultants to interview dozens of application owners, reverse-engineer dependency chains, and make disposition recommendations based on incomplete information. At 2,500 VMs, even the best consulting team would produce inconsistent results — different analysts making different judgment calls across hundreds of applications.
The real risk was not missing the deadline. It was making the wrong modernization decisions at scale — migrating applications that should have been retired, or rehosting applications that needed refactoring.
How CHAI Changed the Equation
Discovery with CHAI Universe
CHAI Universe was deployed agentlessly across the provider's infrastructure. Within 72 hours, it had completed a full scan of all 2,500 VMs — identifying every running application, mapping network connections, cataloging software dependencies, and building a complete topology of the environment.
The discovery process uncovered 340 applications that were not tracked in the provider's CMDB, including 28 that were actively serving production traffic. It also identified 180 VMs running applications that had been decommissioned on paper but never actually shut down — consuming resources and generating costs with no business value.
Assessment with CHAI DART
With the application landscape fully mapped, CHAI DART performed deep analysis of every application. For each one, DART evaluated:
- Code complexity: Language, framework versions, and architectural patterns
- Dependencies: Database connections, API calls, shared services, and file system dependencies
- Cloud readiness: Statelessness, configuration management, logging patterns, and 12-factor compliance
- Business criticality: Traffic patterns, SLA requirements, and integration points
DART generated a 7R disposition for each application — retain, retire, rehost, relocate, repurchase, replatform, or refactor — along with effort estimates, risk scores, and recommended migration wave assignments.
Execution with CHAI Flow
For applications tagged for containerization, CHAI Flow automated the transformation. It generated optimized Dockerfiles, created Kubernetes deployment manifests, and built CI/CD pipeline configurations. Applications that required refactoring received detailed decomposition plans with identified service boundaries.
The provider began executing migrations in parallel with the assessment. By the time DART completed its analysis of the full portfolio, the first wave of 200 applications was already running in the target cloud environment.
The Outcome
The results exceeded every expectation:
- Full assessment in 3 months: All 2,500 VMs and 1,300+ applications assessed with disposition recommendations, effort estimates, and migration wave plans
- 50% infrastructure cost reduction: Achieved through right-sizing, retirement of unused resources, and containerization of overprovisioned VMs
- 340 shadow applications discovered: Brought under management for the first time
- 180 zombie VMs identified and retired: Immediate cost savings with zero business impact
- Consistent, auditable decisions: Every disposition recommendation backed by data, not consultant judgment
The provider completed their datacenter exit on schedule. More importantly, they gained a comprehensive, continuously updated view of their application portfolio — a foundation for ongoing modernization as business needs evolve.