Coupang System Design: Large Video Upload (Backend)

You are asked to design the backend for uploading very large videos (multiple GB) to a platform similar to YouTube. The focus is on the core upload flow from the client selecting a file through to the file being durably stored and initial validation/processing triggered.

Core content

You should describe how the system generates a unique upload ID, accepts chunked uploads, tracks progress, and supports resumable transfers after network interruptions. Explain short-term chunk buffering, how/when chunks are assembled or referenced in object storage, and how the system triggers initial processing tasks such as format validation and metadata extraction once all chunks arrive.

High-level flow/stages

• Upload initiation: auth, create upload session ID, client gets chunk size and endpoints.
• Chunked upload: client uploads chunks (concurrent/parallel), server acknowledges each chunk and updates progress.
• Resumption and recovery: how you record uploaded chunks and allow resume from last successful chunk after failures.
• Finalization: verify integrity, move data from temporary to persistent object storage, trigger processing queue.

Skill signals

You should demonstrate understanding of object storage semantics, data replication and consistency, resumable-upload protocols (e.g., byte-range or session-based), progress reporting, cost trade-offs (temporary vs. long-term storage), scalability patterns (sharding, stateless upload proxies, CDNs), and fault-tolerant designs (checksums, retry/backoff, idempotency).

When preparing, be ready to discuss trade-offs (latency vs cost, single-pass assembly vs reference-joining) and how you'd instrument and test recovery paths.