Types of caching: every layer from browser to database
Browser, CDN, reverse proxy, app memory, Redis and database: what each caching layer stores, who shares it, and how one results-day request moves through them.
Engineering at Scale
Building for 100,000 students at once
37 articles
Articles
How autoscaling works, with AWS Auto Scaling groups
Autoscaling is a feedback loop of metrics, policies and limits. How AWS Auto Scaling groups use target tracking, warm-up, warm pools and health checks, and what doesn't scale.
What is connection pooling, and how does it work?
Why opening database connections is expensive, how pools lend and reuse them, two-tier pooling on large systems, Django's pool options, and the settings that matter.
Point-in-time recovery in Postgres: WAL, backups, restores
Restore Postgres to the second before a mistake: base backups and WAL archiving, recovery settings, pgBackRest, WAL-G and managed services, retention, and restore testing.
How to scale a video learning platform
From transcoding and CDNs to live classes and test-day spikes: an engineer's guide to scaling video learning for thousands of students at once.
How to handle 100,000 concurrent users
Load balancing, caching, queues, read replicas, connection pooling, autoscaling and load testing: how to build a platform that survives its busiest minute.
Horizontal vs vertical scaling: which should you use?
Scale up or scale out? The trade-offs in cost, limits and downtime, what stateless apps need, and why databases usually grow vertically first, with a worked cost example.
Kubernetes HPA: horizontal pod autoscaling explained
How the HorizontalPodAutoscaler calculates replicas, why CPU is often the wrong metric, custom metrics, HPA vs VPA vs KEDA, node autoscaling and tuning.
PgBouncer explained: pooling modes, settings and pitfalls
Session, transaction and statement pooling, what breaks in transaction mode, key pgbouncer.ini settings, PgBouncer vs Pgpool-II and managed proxies, and production tips.
Predictive and scheduled scaling for exam-day traffic spikes
Reactive autoscaling is minutes too late for a 10 a.m. mock test. How to use scheduled actions, predictive scaling and warm pools, and pre-scale databases and caches.
Connection pool size: how to calculate it with Little's Law
Size database pools, Gunicorn workers and pods from real numbers with Little's Law, see why bigger pools slow Postgres down, and follow a worked exam-day example.
Postgres too many connections error: causes and fixes
What 'sorry, too many clients already' means, how to find the culprits in pg_stat_activity, quick relief during an incident, and why raising max_connections isn't the fix.
Upclass for institutes
Launch your academy. Keep every course safe.
Your courses, live classes and tests in your own branded app, with piracy protection on every lecture. Plans start at ₹1,999/month.