OTT streaming · Scale · 2026 — now
From falling over at 40k viewers to 2 lakh+
PlayboxTV’s streaming backend went down at around 40,000 concurrent users. I worked on making it stable, robust and scalable — it now handles over 2 lakh concurrent users firmly.
Visit playboxtv.in- Client
- PlayboxTV — OTT streaming
- Role
- Tech Lead — stability & scale
- Stack
- NestJS · TypeScript · Node.js · MongoDB · Redis + Lua · AWS · Elastic APM
Simplified architecture sketch
- 40k → 2L+concurrent users, stable
- 5×more headroom at peak
- 4-stepweighted cutover: 10 → 25 → 50 → 100%
Context
PlayboxTV streams live TV, sports and on-demand content to phones, browsers and smart TVs. Its core APIs were Express monoliths, and big live events pushed them over at around 40k concurrent users — a blank screen for everyone at once.
The challenge
- 01Stop the outages at peak without pausing the product.
- 02Replace the services that start every stream, without a big-bang release.
- 03Keep partner logins (Hotstar, Jio SSO, FanCode) and smart-TV apps working throughout.
Approach
Fix the hot path
Playback, user and continue-watching moved to NestJS services with caching on Redis, so peak reads stop hitting the database.
Strangle, don’t rewrite
Traffic moved endpoint by endpoint in weighted steps (10 → 25 → 50 → 100%), with load and soak tests gating each step.
Fail safe by default
An edge layer with JWT and rate limits, and rollout flags that fail open, so a cache problem never blocks playback.
See it, then scale it
Elastic APM traces, CodeDeploy pipelines and secrets in AWS Secrets Manager.
Result
- ✦Handles 2 lakh+ concurrent users firmly — up from failing at around 40k.
- ✦Every migrated path has a rollback switch; smart-TV and partner logins kept working throughout.
Team effort on a private codebase. Details in interviews.