Practice: Design a File Converter
File conversion service inspired by CloudConvert and Smallpdf: uploads go to object storage, conversion jobs run on workers from a queue, and results are downloaded from the edge.
Build this system yourself on the canvas below. Add components, connect them, and run traffic simulations to see where your design holds up and where it breaks.
Functional requirements
- Upload a file (up to 100 MB) and choose the output format; large files go straight to object storage
- Convert the file in the background using workers that pick jobs from a queue
- Show the job's status (queued, converting, done, failed) and let the user download the result when done
- Retry a failed conversion automatically, and show a clear error if the file cannot be converted
- Delete both the source and the result 24 hours after the job completes
- Limit how many uploads one user or IP can start per minute, to protect the workers
Non-functional requirements
- Upload accepted < 500ms p99 for the request itself (the file transfer goes to storage directly)
- Status check < 100ms p99, served from the status cache; 10K status checks and downloads per second at peak
- A small file (under 5 MB) converts within 30 seconds of upload
- 99.9% of jobs finish successfully; failed jobs retry up to 3 times with backoff before being marked failed
- Workers scale with queue depth, so a burst of 200 uploads per second waits instead of failing
- Results are served from the CDN; nothing is kept longer than 24 hours, so storage cost stays flat
- 99.9% availability for upload and status