Requirements

Functional Requirements:

• Readers can search and read eBooks in the browser.
• Authors can submit their book to a distributor's account.
• Distributors can read the books sent to their account and enable them for distribution.

Non-Functional Requirements:

• A book can be opened for reading in less than 3 seconds, and each page turn takes less than 100ms.
• Storage of books is 99.99% durable.
• Availability is 99.9%.
• Supports 10M DAU each viewing 5 book metadata and doing 20 page turns per day
  • 500 metadata/s
  • 2000 page turns/s

API Design

• Reader's browser will use websocket to stream ebook.
• Authors's browser will use HTTP POST to upload ebook.
• Distributor APIs will be REST APIs.
  • GET /books
  • GET /books?id={id}
  • PATCH /book?id={id} with body {enabled: true}

High-Level Design

Books will be stored in unstructured low-cost object storage like S3. Metadata about books will be stored in a database. There will be a cache for the metadata. If searching directly on the database adds too much load, we can have a dedicated elasticsearch service for searching for books. There will be a web tier to serve the API. Behind that there will be a BookService that is responsible for loading books from S3 and streaming them over the websocket.

Detailed Component Design

Deep dive into 2-3 key components. Explain how they work, how they scale, discuss tradeoffs, capacity, and any relevant algorithms or data structures.

Database: Use a relational database since it is mostly read traffic, we may want to do JOINs for the different usecases, and the schema is mostly known and doesn't need extreme flexibility. It will have a books table that has columns for author id and publisher id, plus book metadata columns. It will have an authors table and a publishers table.

Cache: Requests to get a book metadata by ID will hit the cache first, which is Redis. If it misses, it retrieves the metadata from the database and stores it in the cache.

BookService: A stateless replicated microservice that receives gRPC requests from the web tier. On a request for streaming a book, it downloads the book from S3 to local disk and streams it to the client over websocket.