Telecom Argentina
An OTT platform that worked every day of the year except the ones that mattered most.
Stack
- Microservices
- Event-driven architecture
- Distributed systems
- AWS
- Cloud migration
- OTT streaming
Problem
The OTT streaming platform ran on-premises and held up fine under ordinary load. Ordinary load was not the problem. The platform's worst hour and its most valuable hour were the same hour: a major football match, where every subscriber in the country arrives inside the same minute and stays for ninety. Latency climbed exactly when the audience was largest and least forgiving, and the bottlenecks were structural — an origin and a delivery tier that could not be scaled independently of each other, sized for an average that the business did not actually care about.
Approach
This was a consultancy and pre-sales engagement, so I ran an end-to-end assessment rather than starting with a proposal: the platform architecture, the software delivery lifecycle, and the business-critical workflows around live events. Reading the delivery pipeline mattered as much as reading the topology — a platform that cannot ship a fix during a match has a delivery problem wearing an architecture costume. The recommendation was a cloud-native migration to AWS with distributed microservices and event-driven patterns, sequenced so the pieces carrying live traffic moved first and the rest followed on evidence.
The tradeoff
The appetite in the room was for a rebuild — a clean platform, designed for the peak, cut over at once. I argued against it, and the argument cost me some enthusiasm. A big-bang cutover would have put the delivery date of a system whose whole purpose is a fixed calendar of live events on a hypothesis. The phased migration is slower, and it means living with two platforms and an integration seam for longer than anyone wants. It also means the season does not depend on the rewrite landing.
Impact
Assessment approved · the next phase of the transformation funded
- Architectural assessment of the on-premises OTT platform delivered end to end — topology, delivery lifecycle, and the business-critical live-event workflows
- Scalability, latency, and operational bottlenecks affecting live content delivery identified and traced to their structural causes
- Cloud-native modernization strategy designed — on-premises to AWS, distributed microservices, event-driven patterns
- Architectural patterns proposed to reduce end-to-end streaming latency under massive concurrent traffic
- Architectural vision and roadmap presented to client stakeholders, securing approval for the next phase — a consultancy and pre-sales assessment, not a platform I went on to build
What I learned
The deliverable here was a decision, not a system. That took me a while to say without apologizing for it. This was consultancy and pre-sales: I delivered the assessment, the architecture, and the roadmap, and handed them over — the build that followed was not mine. An assessment that ends in an approved roadmap has done its whole job; the work is in being right about where the platform breaks and honest about what fixing it costs. I also learned to look at the release process before the architecture diagram. The diagram tells you how the system was intended to work. The release process tells you how it actually does.