System design field notes

Size the system
before you draw it.

A quick, inspectable back-of-the-envelope estimate for request load, peak capacity, compute, data transfer, and optional storage growth.

01 / Inputs

Traffic assumptions

editable

Daily active users DAU

users

Approximately 500 million Requests per user per day average

req/user

Peak traffic multiplier 3× Accounts for daily bursts instead of sizing only for the 24-hour average.

Server capacity

RPS

Target utilisation

Leaves runtime headroom; failover capacity is considered separately.

Extended metrics

Average request size

Average response size

02 / Estimate

Capacity snapshot

Peak-aware

Baseline service instances

—

Requests / day

—

DAU × requests per user

Average requests / sec

—

Requests / 86,400 seconds

Peak requests / sec

—

Effective RPS / server

—

Daily data transfer

—

request + response payload

Peak network throughput

—

payload only; excludes protocol overhead

Throughput baseline only. In an interview, add availability capacity explicitly—for example N+1, one-zone loss, or a stated failover policy.

03 / Optional module

Storage footprint

Use this when retained data materially shapes the design—for example media, documents, events, backups, or long-lived user content.

Open only when relevant

Estimate daily growth and retained capacity Items × average payload, adjusted for metadata, retention, and durability copies.

New items per day

items

Average item size

Metadata / derivative

KB/item

Retention period

days

Durability copies / overhead 3× Use 1× for raw data, or approximate replication / erasure-coding overhead when implementation details are unknown.

Retained storage footprint

—

Raw growth / day

—

single logical copy

Provisioned growth / day

—

Provisioned growth / 30 days

—

A planning shortcut; excludes deletion, compression, and lifecycle tiering.

04 / Reference

Useful orders of magnitude

Conversation starters, not guarantees. Real capacity depends on query shape, indexes, hardware, payloads, latency targets, and cache hit rate. Benchmark your own path before committing.

ComponentTypical BOTEC rangeRule of thumb

Relational database10²–10⁴ QPS~1K QPS baseline

Key-value store10⁴–10⁵ QPS~10K QPS baseline

In-memory cache10⁵–10⁶ ops/s~100K ops/s baseline

Network latency0.5–150 mssame DC → inter-region

Monthly volumedaily × 30fast planning shortcut

05 / Engineering Manager lens

Lead the conversation, not every implementation detail.

Use the estimates to expose constraints and trade-offs. Go deeper only where scale, risk, or the interviewer makes it relevant.

Frame

Clarify users, critical journeys, SLOs, constraints, and what is out of scope.

Estimate

Use round numbers to find the likely bottleneck—not to manufacture precision.

Design

Draw the simplest end-to-end path, then scale only the pressured components.

Operate

Cover failure modes, observability, security, cost, and safe rollout at the right depth.

Decide

State trade-offs, ownership boundaries, and what evidence would change the decision.

06 / Drawing kit

Core system design icons for your whiteboard.

A deliberately simple canvas with the essential interview components, including single-instance and scaled-out variants. Choose an editable Excalidraw sheet or Google Drawings-ready PNG assets.

Download Excalidraw icon sheet Download Google Drawings pack

20 grouped components Excalidraw or Google Drawings.
Just copy-ready icons.