Katkat
Back

Design homepage viewport rendering with deduplication

NetflixSoftware Engineer · Technical Screen
MediumSoftware EngineerTechnical ScreenSystem Design

Last updated

This problem tests the ability to design a content-delivery service that optimizes viewport rendering through client-side requests, server-side deduplication, and efficient data retrieval patterns.

Problem Statement

You are tasked with designing the backend service for a streaming platform's homepage. The homepage consists of several horizontal "rails" (e.g., "Continue Watching," "Trending Now," "Top 10").

Requirements

  1. Viewport-based Fetching: The client only requests data for rails currently in the viewport to minimize payload size and latency.
  2. Deduplication: A specific movie or show might exist in multiple lists (e.g., it appears in both "Trending" and "Action Movies"). The client must not display the same item twice; the service must ensure that items returned in the current batch do not repeat IDs already rendered on the page.
  3. Pagination: Each rail has a large number of items. The system must support loading more items within a rail without re-fetching existing data.
  4. Performance: The service must handle high QPS with sub-100ms latency.

Constraints & Assumptions

  • Assume 50 million DAU and 1,000 requests per second at peak.
  • Personalization (ranking) happens in real-time based on user history.
  • The service should return a JSON response containing a list of rails and the unique items within them.

Deliverables

  • Define the API endpoints for requesting homepage content.
  • Propose a data model for storing content rankings and metadata.
  • Describe the service architecture to handle deduplication logic efficiently.
  • Explain how to handle caching and personalization state.

Hints

Hint 1

Consider passing a 'seen_item_ids' set or a Bloom filter in the client request to allow the server to filter out duplicates during the ranking phase.

Hint 2

Use a "Backends for Frontends" (BFF) pattern to aggregate data from multiple microservices and perform the final deduplication before sending the response to the client.

Hint 3

Implement a two-tier caching strategy: a global cache for generic list metadata and a per-user cache for personalized ranking orders.

Follow-up questions

  • How would you handle the 'cold start' problem for a new user where no personalization data exists?
  • If a user scrolls rapidly, how would you design a client-side pre-fetching strategy to minimize perceived latency?
  • How would you instrument the system to measure the impact of your deduplication logic on user engagement (e.g., CTR or watch time)?