How AWS Architecture Handled 1.5M Requests in One Hour for a Flash-Seat Exam Booking

·78 views

A national certification exam with 20,700 seats across five levels needed a new registration system after its previous platform crashed and double-booked seats when registration opened. The redesigned system on AWS uses an atomic Redis script as the sole arbiter for seat allocation, ensuring no seat is ever sold twice and no database is involved in the admission decision. Static content was served via CloudFront and S3, with an AWS WAF web ACL handling roughly three in four requests without reaching origin servers. During the first hour of the live event, the system served over 1.5 million origin requests, peaked at 54,800 requests per minute, maintained median latency under 10 milliseconds, and recorded a server error rate below 0.01%. This is the first installment of a three-part series detailing the architecture, edge strategy, and capacity planning behind the solution.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Newsletter subscribe forms went dead after account was disabled, unnoticed for weeks

The author of a Buttondown-hosted newsletter discovered that the account had been disabled, with notices dated 19 and 24 October, after opening the dashboard for another reason. Fetching the public newsletter URLs returned 404 responses, meaning the subscribe forms embedded in 26 posts had been submitting to nothing. A monitoring script run 29 times between 18 October and 10 November logged a subscriber count of zero each time, but it queried the API rather than checking whether readers could actually reach the signup page, so it never flagged the problem. The API continued to function on the disabled account, which made the zero reading look like an ordinary result rather than a failure. The Wayback Machine shows the public page was still live on 5 October, so the outage window can only be bounded to somewhere in the 35 days before discovery.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

PulseROI Redesigns Influencer Campaign Monitoring to Run Reliably on Serverless

Indian direct-to-consumer brands spend heavily on YouTube creator campaigns but typically stop tracking performance once videos go live, even though purchases can follow weeks or months later. While building PulseROI, a tool for continuous campaign monitoring, the developer found that a naive synchronous loop hit Vercel's serverless timeout limits once campaigns exceeded about 15 videos. The fix was to restructure the pipeline into batched, asynchronous processing that tolerates individual API failures using Promise.allSettled. A 25-second timeout guard was also added so the cron job exits cleanly before Vercel terminates it.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

WAV, MP3, FLAC and AAC Compared: A 60-Second Round-Trip Test of Audio Quality Loss

A client asked for a 90 MB WAV demo to be shrunk, prompting an experiment with a synthesized 60-second 48 kHz, 16-bit stereo reference track. The track was converted across several formats using ffmpeg 8.0 and decoded back to measure file size and signal-to-noise ratio against the original. Results showed FLAC preserved the audio bit-exactly at about 31% of the WAV size, while lossy MP3 and AAC files were far smaller but lost fidelity. Re-encoding lossy files compounded the losses, with a two-step chain scoring 6.4 dB lower than a single encode. The author recommends FLAC for archiving and MP3 320k or AAC 256k for sharing, and advises against lossy-to-lossless upgrades or lossy-to-lossy chains.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

GPT Image 2.6 Edits Can Be Made Auditable With Change Contracts

A developer guide argues that precise image editing with GPT Image 2.6 should be governed by explicit rules about which pixels may change and what must stay intact, rather than relying on prompts alone. It proposes pairing natural-language instructions with a transparent PNG annotation layer and checking results against preservation rules. The article describes server-side routing that maps business intents such as preview or final to allow-listed model variants, Flare for rapid iteration and Sunburst for detail-focused finals. It also recommends structured change contracts with source and markup hashes, plus local validation of annotation files against the API's dimension and size limits. Production guidance additionally covers secure remote input fetching and recording usage rights.

0 comments Read more at DEV Community