Let’s play a game: Try to design a multi-exabyte storage system. It doesn’t need to be the most throughput or latency efficient, but just something that works. It’s already not a trivial problem, and it could actually be a very interesting interview question and discussion.
Well, once you have thought about how would you do it, you can read this article to learn how Dropbox actually did it and it is using in production.
Originally Dropbox has been storing its data in AWS S3, and a few years ago they moved everything to an in-house storage that they designed called “Magic Pocket”. This article is an explanation of the design of the system and how all the pieces are working together, to let the user store and retrieve data from Dropbox.
I have read and studies a lot of storage solutions, but this is one of the most complete write-up I have read in a while, considering also that this is not a research paper, this is actually used in a multi-exabyte production environment.