Technical case study

Redis-first exam flow

Optimizing question reads and autosave hot paths while preserving durable persistence.

LaravelRedis HashQueueS3k6Distributed lock

01 · Context

The exam system needs low-latency question reads and answer autosaves. The previous S3-first hot path increased response time as concurrent participation grew.

02 · Problem

Reduce latency for getQuestions and save, prevent duplicate exam creation, and retain a reliable fallback and durable persistence path.

03 · Architecture

Processing flow

01 Exam request
02 Distributed lock
03 Redis Hash
04 Persist queue
05 S3 storage

04 · Result

In a controlled 25-VU benchmark, throughput improved by about 6.1×; getQuestions p95 dropped by roughly 69% and save p95 by roughly 70% versus the S3-first flow.

05 · Ownership

I directly designed, implemented and benchmarked the Redis-first flow.

06 · Trade-off

A faster Redis hot path requires explicit fallback, rebuild and persistence strategies. Queue failures need retries and monitoring, and benchmark numbers must not be presented as production SLAs.

Want to go deeper?

I can walk through the technical decisions, constraints and alternatives in a conversation.

Email me