# Design Zoom: system design interview question

> Live group calls require low delay, bandwidth control, enforced participant permissions, and recovery after regional failures.

Source: https://sysround.com/questions/design-zoom
Updated: 2026-11-05


Difficulty: hard. Levels: Mid-level, Senior, Staff+.

## How the interviewer opens it

"Design a video conferencing service with meeting rooms and participant controls. Severe bandwidth limits may require audio-only calls. A regional failure may briefly interrupt media, but the meeting must recover."

## What to clarify first

- Scale
- Room access and controls
- Media volume
- Read-to-write mix
- Latency
- Clients and networks
- Consistency and availability
- Regional failures
- Durability and privacy
- Scope limits

## Scope of a strong answer

Core requirements:

- Create, join, and leave private meeting rooms.
- Send live audio and video in rooms with up to 100 participants.
- Let hosts mute and remove participants.
- Adapt media quality to bandwidth limits and restore calls after regional failures.

Usually out of scope:

- Record calls or produce transcripts.
- Support broadcasts with thousands of viewers.
- Provide strict end-to-end encryption.
- Build payment or subscription services.

## What interviewers probe

### Media paths

A room can contain 100 participants. A poor media path can exhaust a client's upload capacity or add too much delay.

### Audio and video quality

Each participant has different network limits. Delayed feedback and sudden bitrate changes can damage audio before the system reduces video.

### Participant controls

A host's mute or removal must change the media path. Concurrent controls and reconnects must not restore a participant's old rights.

### Regional failures

A failed region can lose every active media session in that region. Reconnects can overload another region and revive old room state.

## Common mistakes

- The design sends every video to every participant despite the nine-video display limit.
- The design treats an interactive call like a video file stream.
- The design uses one bitrate for every receiver.
- The server trusts the client to enforce host mute and removal.
- The design assumes DNS changes preserve active media sessions.
- The capacity estimate excludes TURN traffic and spare capacity for a regional failure.

## The bar by level

- **Mid-level:** Separates control and media paths. Explains a working group call, connectivity fallback, basic bandwidth reduction, and server-enforced host controls. Recognizes that regional recovery requires client reconnects.
- **Senior:** Designs per-receiver media selection and congestion response. Handles repeated controls, concurrent changes, and removal during reconnect. Defines regional recovery with durable permissions and spare capacity.
- **Staff+:** Quantifies network throughput, packet rates, TURN traffic, and recovery limits. Proves that isolated workers cannot retain authority indefinitely. Budgets the recovery target and explains cross-continent delay limits.

Practice this question with an AI interviewer: https://sysround.com/practice/new?problem=video-conferencing


---
SysRound (https://sysround.com) is an AI practice tool for system design interviews. It has a voice or typed AI interviewer and a shared whiteboard. At the end, you get feedback on each skill and a model answer. The AI feedback can be wrong.