KIRTAN PARIKH

Software & DevOps Engineering
PAGE LOADING
Back to Portfolio
Case Study

Scaling Real-Time Support Infrastructure

Architecting a high-availability Socket.IO gateway on AWS for enterprise clients.

Node.js
AWS EC2
S3
PostgreSQL

Professional Note

This case study reviews architectural challenges solved during my internship. To comply with NDA, all proprietary data, client details, and internal IPs have been obfuscated. The demo video uses a sanitized 'clean-room' environment.

Deep Dive: Resilient File Transfer Protocol

S3 Architecture that bypasses server bottlenecks and self-heals expired links

Phase 1: Zero-Trust Direct Uploads

To prevent server bottlenecks, I implemented a direct-to-cloud pattern. Clients request a temporary AWS S3 Presigned URL from the backend, then upload binary data directly to a private bucket. This bypasses the Node.js event loop, ensuring chat latency remains <100ms even during large file transfers.

Impact: Eliminated server memory spikes during concurrent uploads

S3 Presigned Upload Sequence

Diagram: S3 Presigned Upload Sequence

403 Expiry Interception & Refresh Flow

Diagram: 403 Expiry Interception & Refresh Flow

Phase 2: Self-Healing Link Recovery

Secure read-tokens expire after 24 hours, causing broken images. I implemented a frontend interceptor that detects 403 Forbidden responses from S3. It silently requests a fresh token from the API and retries the image load transparently, creating a seamless user experience.

Result: Zero user-visible errors from expired presigned URLs

The Architectural Shift

Evolution from resource bottleneck to horizontally scalable infrastructure

Phase 1: Decentralized Mesh

Initial prototype where every client website maintained a direct, heavy socket connection to the dashboard. Scalability bottleneck: O(N) connections per agent.

Phase 1: Decentralized Mesh Architecture

Problem: Each new client multiplied server load exponentially

AWS Cloud Topology

Production-grade infrastructure spanning compute, persistence, storage, and identity layers

AWS Cloud Architecture Diagram

Production Environment: EC2 (App Layer), RDS (Persistence), S3 (Storage), Cognito (Identity)

Live System Demo

Watch the full system in action — sanitized for public viewing

Sanitized Technical Demo (Clean-room environment)

Production Outcome

99.9%
Uptime
<100ms
Latency

Deployed to production achieving enterprise-grade reliability and performance benchmarks