How the interviewer opens it
“Design a global network that caches and delivers media near viewers. It must handle popular releases while it limits origin traffic and delivery costs.”
The prompt is vague on purpose. The first minutes are for you. Reduce the prompt to the core features and agree on the numbers before you draw.
What to clarify first
- Media format
- Viewer scale
- Catalogue and cache capacity
- Read and write mix
- Latency
- Consistency and availability
- Origin capacity and cost
- Durability
- Access and exclusions
The interviewer has the answers. In a SysRound round, Alex gives these answers, as a real interviewer does.
Scope of a strong answer
Core requirements
- Serve media from nearby, healthy edge sites.
- Handle popular releases without excess origin load.
- Control cache use and delivery costs.
- Serve complete segments and valid byte ranges.
Usually out of scope
- Live video
- Media encoding and upload workflows
- Digital rights management
- A custom Internet backbone
- Billing
What interviewers ask about
- Popular releases and origin load
- Viewers can request the same new segments before any edge has them. Independent misses can exceed origin capacity despite a high average hit rate.
- Site choice and failures
- The nearest site can lack capacity or cached media. Viewers can continue attempts to reach a failed site after the network selects an alternative.
- Media requests and cache space
- The catalogue exceeds cache capacity, and demand varies by region. Byte ranges and object sizes can make request hit rates hide high origin traffic.
Common mistakes
- The design treats the network as a metadata service instead of a system that delivers media bytes.
- The capacity estimate counts requests but ignores bandwidth and cache space.
- Each cold request reaches the origin during a popular release.
- The design assumes every edge can store the full catalogue.
- The design ignores byte ranges, incomplete fills, or manifest freshness.
- The cost estimate treats a cache hit as free delivery and ignores paid transit.
The bar by level
- Senior (L5/E5)
- Designs global site selection and shared miss control. Protects origin capacity during releases and failures. Connects byte hit rates to cache policy and cost.
- Staff (L6+/E6+)
- Sets explicit overload behavior and regional failure budgets. Quantifies spare capacity, cold recovery, and transit trade-offs. Separates control failures from delivery failures.