𝗛𝗼𝘄 𝘁𝗼 𝗮𝗻𝘀𝘄𝗲𝗿 𝘀𝘆𝘀𝘁𝗲𝗺 𝗱𝗲𝘀𝗶𝗴𝗻 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻𝘀 𝘂𝘀𝗶𝗻𝗴 𝘁𝗵𝗲 𝗦𝘆𝘀𝘁𝗲𝗺 𝗗𝗲𝘀𝗶𝗴𝗻 𝗠𝗮𝘀𝘁𝗲𝗿 𝗧𝗲𝗺𝗽𝗹𝗮𝘁𝗲?

Here is the System Design Master Template, which you can use as a basis for any system design problem you may encounter during an interview.

A list of topics that you should know:

𝟭. 𝗟𝗼𝗮𝗱 𝗕𝗮𝗹𝗮𝗻𝗰𝗲𝗿. Distributes inbound requests across healthy nodes using rules such as round-robin or least-connections. Keeps latency stable and enables rolling upgrades without downtime.

𝟮. 𝗔𝗣𝗜 𝗚𝗮𝘁𝗲𝘄𝗮𝘆. Single entry that authenticates callers, enforces quotas, and routes each call to the right microservice. It can also aggregate responses so that clients hit a single endpoint.

𝟯. 𝗖𝗼𝗻𝘁𝗲𝗻𝘁 𝗗𝗲𝗹𝗶𝘃𝗲𝗿𝘆 𝗡𝗲𝘁𝘄𝗼𝗿𝗸 (𝗖𝗗𝗡). An edge network that caches immutable assets near users, cutting round-trips and shielding origins from spikes. Often handles TLS and compression as well.

𝟰. 𝗠𝗲𝘁𝗮𝗱𝗮𝘁𝗮 𝗦𝗲𝗿𝘃𝗲𝗿. Stores pointers, permissions, and object properties in a small, fast store. Replicates data for reads and writes to block servers when files are changed.

𝟱. 𝗕𝗹𝗼𝗰𝗸 𝗦𝗲𝗿𝘃𝗲𝗿. Persists data chunks on commodity disks, replicates across racks, and self-heals after failures. Clients stream blocks in parallel for throughput.

𝟲. 𝗦𝗵𝗮𝗿𝗱 / 𝗣𝗮𝗿𝘁𝗶𝘁𝗶𝗼𝗻 𝗠𝗮𝗻𝗮𝗴𝗲𝗺𝗲𝗻𝘁. Splits a growing dataset by user ID, time, or hash so that each DB instance remains small. Rebalances shards when a node fills or a hot key appears.

𝟳. 𝗖𝗮𝗰𝗵𝗲 (𝗥𝗲𝗱𝗶𝘀/𝗠𝗲𝗺𝗰𝗮𝗰𝗵𝗲𝗱). Holds hot keys in RAM with TTL, counters, and locks. Reduces read latency from milliseconds to microseconds and offloads the primary store.

𝟴. 𝗗𝗶𝘀𝘁𝗿𝗶𝗯𝘂𝘁𝗲𝗱 𝗠𝗲𝘀𝘀𝗮𝗴𝗶𝗻𝗴 (𝗞𝗮𝗳𝗸𝗮/𝗥𝗮𝗯𝗯𝗶𝘁𝗠𝗤). Durable log that decouples producers from consumers, provides replay, back-pressure, and at-least-once delivery. Enables event-driven designs.

𝟵. 𝗡𝗼𝘁𝗶𝗳𝗶𝗰𝗮𝘁𝗶𝗼𝗻 𝗦𝗲𝗿𝘃𝗶𝗰𝗲. Pulls events from the queue, batches them, and pushes email, SMS, or mobile alerts. Retries with exponential back-off and user-level throttling.

𝟭𝟬. 𝗝𝗼𝗯 𝗪𝗼𝗿𝗸𝗲𝗿𝘀. Stateless containers that pick tasks such as video encoding or thumbnail generation from a queue. Scale horizontally by adding replicas.

𝟭𝟭. 𝗗𝗶𝘀𝘁𝗿𝗶𝗯𝘂𝘁𝗲𝗱 𝗟𝗼𝗴𝗴𝗶𝗻𝗴 𝗮𝗻𝗱 𝗧𝗿𝗮𝗰𝗶𝗻𝗴. Ships structured logs and distributed spans to a central store, then indexes by request ID. Let's you trace a single user action across dozens of services.

𝟭𝟮. 𝗕𝗮𝘁𝗰𝗵 / 𝗦𝘁𝗿𝗲𝗮𝗺 𝗣𝗿𝗼𝗰𝗲𝘀𝘀𝗶𝗻𝗴 (𝗦𝗽𝗮𝗿𝗸, 𝗛𝗮𝗱𝗼𝗼𝗽). Transforms raw events into aggregates, ML features, and reports. Supports windowed joins for near-real-time dashboards.

𝟭𝟮. 𝗠𝗲𝘁𝗿𝗶𝗰𝘀 𝗮𝗻𝗱 𝗔𝗹𝗲𝗿𝘁𝗶𝗻𝗴. Scrapes time-series data, stores it efficiently, and renders dashboards. The Alert Rules page alerts you when latency, error rate, or saturation crosses SLOs.