Design Durable, Cost-Efficient Storage and Delivery for Very Large Video Files

Quick Overview

A technical product manager question on designing storage for very large video files. It tests whether a candidate can set requirements and estimates, design resumable uploads, durable and cost-tiered storage and a playback path, and explain trade-offs such as replication versus erasure coding.

Design Durable, Cost-Efficient Storage and Delivery for Very Large Video Files

Company: Google

Role: Product Manager

Category: System Design

Difficulty: medium

Interview Round: Other

Design a system that stores very large video files. Explain how a video gets into the system, how it is kept safe and affordable to store over time, and how it is read back when someone needs it, and call out the main trade-offs you make along the way. ```hint Bytes versus bookkeeping The video data and the information about each video differ enormously in size and in how they are accessed. Decide where each one lives, then ask what should happen when a multi-gigabyte upload breaks near the end. ``` ### Clarifying Questions - Does the system sit behind a public video-sharing product, where popular videos are watched many times, or a backup and archive service, where most files are written once and seldom read? - How large are the biggest files the system must accept, and how many uploads a day should it handle? - Is streaming playback in a browser or app in scope, or only storing and downloading the original files? - How long must videos be kept, and must a deletion request remove every copy, including backups? ### What a Strong Answer Covers - Requirements and scale assumptions stated before the architecture, with a rough estimate of storage growth and upload bandwidth. - An upload path that copes with huge files and unreliable connections and verifies that nothing was corrupted on the way. - A storage design that separates video data from metadata and addresses durability and cost through redundancy choices and storage tiers. - A read path matched to how the videos are used, including delivery to viewers far from where the data is stored. - Failure handling, the key trade-offs, and the measures a product manager would watch. ### Follow-up Questions - Storage costs are growing faster than revenue. Which levers would you pull first, and which of them would users notice? - How would you avoid storing many identical copies of the same video, and what privacy risk does that create? - A user deletes a video. What has to happen across the system, and how would you show the video is really gone?

Overview: A technical product manager question on designing storage for very large video files. It tests whether a candidate can set requirements and estimates, design resumable uploads, durable and cost-tiered storage and a playback path, and explain trade-offs such as replication versus erasure coding.

Google logo
Google
Aug 21, 2025
mediumProduct ManagerOtherSystem Design
0
0

Design a system that stores very large video files. Explain how a video gets into the system, how it is kept safe and affordable to store over time, and how it is read back when someone needs it, and call out the main trade-offs you make along the way.

Clarifying Questions Guidance

  • Does the system sit behind a public video-sharing product, where popular videos are watched many times, or a backup and archive service, where most files are written once and seldom read?
  • How large are the biggest files the system must accept, and how many uploads a day should it handle?
  • Is streaming playback in a browser or app in scope, or only storing and downloading the original files?
  • How long must videos be kept, and must a deletion request remove every copy, including backups?

What a Strong Answer Covers Guidance

  • Requirements and scale assumptions stated before the architecture, with a rough estimate of storage growth and upload bandwidth.
  • An upload path that copes with huge files and unreliable connections and verifies that nothing was corrupted on the way.
  • A storage design that separates video data from metadata and addresses durability and cost through redundancy choices and storage tiers.
  • A read path matched to how the videos are used, including delivery to viewers far from where the data is stored.
  • Failure handling, the key trade-offs, and the measures a product manager would watch.

Follow-up Questions Guidance

  • Storage costs are growing faster than revenue. Which levers would you pull first, and which of them would users notice?
  • How would you avoid storing many identical copies of the same video, and what privacy risk does that create?
  • A user deletes a video. What has to happen across the system, and how would you show the video is really gone?