Back to problems

S3-Style Storage with Dedup

System Design · Snowflake · Medium

Requirements Users can upload files of any size, and the service must return a durable object identifier. Minimizing storage spend is the primary goal: files with exactly matching contents must occupy physical storage a single time, even when referenced by many logical names. Clients must still retrieve and remove objects through their logical paths, without being exposed to the deduplication mechanism. Complete S3 functionality—including object versioning, per-object ACL…

Checking your access…