# Scalability

Scalability is a system’s ability to support growing workload while maintaining agreed performance and acceptable resource costs.

- Type: principle
- Domain: Engineering & architecture
- Level: Intermediate
- Last reviewed: 2026-10-19
- Original: [Soroush’s Product Wiki](<https://wiki.soroush.uk/topic/scalability/>)

## What it means

Scalability describes how system behavior changes as demand, data, users, or geographic reach increase. A scalable design can accommodate a defined growth dimension through appropriate resources or structural changes. It is different from being fast at today’s load and from elasticity, which concerns adjusting capacity as demand changes.

## Why it matters

Growth can expose bottlenecks that small workloads never reach. Explicit scalability expectations help teams avoid both premature distributed complexity and a product architecture that cannot support credible demand.

## When to use it

Use it when planning launches, large tenants, data growth, or sustained expansion. Define the workload dimension and acceptable costs before choosing technology. If a simpler system meets foreseeable demand with headroom, complexity for hypothetical scale may be unnecessary.

## How it works

1. Define growth scenarios and user-facing performance targets.
2. Measure current capacity, resource use, and limiting dependencies.
3. Identify which components or data structures constrain growth.
4. Test capacity changes or architecture options against realistic workloads.
5. Track cost per unit of work and re-evaluate as bottlenecks move.

## A practical example

A reporting product handles ten thousand records quickly but slows sharply at one million. Adding application servers does not help because every report scans the same database table. The team changes its query and storage approach, then retests the expected data-growth scenario.

## Where it breaks

No architecture scales without limits or costs. Improving one bottleneck can expose another, and a distributed design adds coordination and operational overhead. Demonstrated capacity under one request mix does not establish capacity under every workload.

## Common mistakes

- Calling a fast small-system benchmark evidence of scalability.
- Adding compute before locating the bottleneck.
- Assuming cloud hosting removes database, quota, or budget limits.

## Sources

1. [Architecture strategies for capacity planning](<https://learn.microsoft.com/en-us/azure/well-architected/performance-efficiency/capacity-planning>) — Microsoft · Official documentation
2. [Architecture strategies for optimizing scaling and partitioning](<https://learn.microsoft.com/en-us/azure/well-architected/performance-efficiency/scale-partition>) — Microsoft · Official documentation

## Related topics

- [Load Testing](<https://wiki.soroush.uk/topic/load-testing/>)
- [Horizontal and Vertical Scaling](<https://wiki.soroush.uk/topic/horizontal-and-vertical-scaling/>)
- [Non-Functional Requirements](<https://wiki.soroush.uk/topic/non-functional-requirements/>)
