System Design AI
All designs

Classic designProduct & backendadvanced

8. Design a Video Streaming Platform

Turn uploads into playable renditions and serve viewers through unreliable networks.

The brief

Design an on-demand video platform from resumable upload through transcoding and playback. Explain when a video becomes playable, how viewers select bitrate and how popular content is delivered. Handle failed processing, deletion and a sudden audience spike without assuming live streaming is in scope.

  • Upload 100,000 videos/day averaging 500 MB and ten minutes of playback each.
  • Serve 1 million concurrent viewers at peak, with average delivered bitrate of 3 Mb/s.
  • Offer three playback renditions. Processing can fail and a new video can become popular immediately.

Constraints

Playback startup target≤ 3 seconds
Declare p95 startup time for a cached video on an adequate network; state excluded network conditions.
Playable publication
Expose only complete manifests and ready segments; represent processing failures honestly.
Deletion and cache policy
Define removal from discovery, authorization and cached delivery with explicit propagation limits.

What to cover

  1. 01

    Upload and processing states

    Show durable upload, job identity, renditions, retries and publication.

  2. 02

    Playback path

    Describe manifests, segmented delivery, bitrate selection and access checks.

  3. 03

    Capacity estimate

    Calculate ingress storage, approximate processing demand and peak egress bandwidth.

  4. 04

    Spike and outage walkthrough

    Trace a failed rendition and a cold popular video; compare origin protection strategies.

Worked designs

Explore the architecture and decisions, then build on an example with Coach.

Review rubric

AI feedback uses these criteria. Scores are practice feedback.

Media lifecycle

Uploads, processing retries and publication have consistent state.

30points

Playback behavior

Adaptive delivery and access rules support the stated experience.

25points

Bandwidth and storage

Unit-aware estimates expose the major costs and limits.

25points

Spike and failure response

Recovery protects the origin and states cache tradeoffs.

20points

Discussion

Share an approach, ask a question, or tag @Coach.

Share your interview experience

Published under your alias as a community report. Leave out interviewer names and confidential material.

Loading discussion…