Engineering a web application for steady-state web traffic is straightforward; engineering an application for sudden, extreme traffic spikes where 50,000 active students interact simultaneously within a 45-minute window requires an entirely different architectural discipline. In this article, we reveal the server optimization, database indexing, and queue scaling techniques engineered by SOFT IT TECHNOLOGIES for Sohoj Pathshala (sohojpathshala.com).
1. The Peak Concurrency Equation
To understand the server strain of a national model test, examine the mathematical reality of HTTP traffic during a synchronized examination:
- 50,000 students active simultaneously on the exam canvas.
- Periodic auto-save pings sent every 60 seconds = ~833 HTTP POST requests/sec continuously.
- Exam timer expiration: 50,000 final submission payloads arriving in a 5-second burst = 10,000 requests/sec.
2. The 4 Golden Rules of Laravel Concurrency Scaling
To withstand this load on cost-efficient cloud infrastructure, our engineering team adhered to 4 non-negotiable architectural rules:
Rule 1: Eliminate Database Queries from Critical Read Paths
Active exam questions and options are pre-rendered into static JSON blobs stored in Redis memory. When examinees start their test, their browser fetches question sets directly from Redis cache with zero SQL read operations.
Rule 2: Push All State Changes to Asynchronous Queues
Student answers are pushed into a dedicated Redis queue. The web controller does not evaluate marks, does not query databases, and does not calculate leaderboards. It merely verifies HMAC request signatures and returns `HTTP 202 Accepted` within 25 milliseconds.
Rule 3: Optimize PHP-FPM and Opcache Configuration
PHP 8.2 Opcache JIT (Just-In-Time) compilation was enabled with `opcache.enable_cli=1` and `opcache.memory_consumption=512`. PHP-FPM process managers were tuned to `pm = static` with 120 dedicated worker processes, eliminating runtime process spawning overhead.
Rule 4: Edge CDN Asset Offloading
All static CSS, JavaScript, WebP images, and fonts are cached permanently on Cloudflare edge servers with immutable cache-control headers, ensuring 98% of asset requests never reach the origin server.
3. Summary
By engineering for concurrency from day one, Sohoj Pathshala achieved what few regional EdTech platforms can claim: 100% operational reliability, instantaneous exam grading, and complete peace of mind for educators and examinees alike.