Series

Scalability

A series on scalability – from individual instances to distributed systems and the organisations that develop them.

  1. Introduction

    Fifty Users

    A portal crashes when there are 50 concurrent users. This marks the start of a series on the topic of scalability – from individual instances through the architecture of distributed systems to the organisation.
  2. Part 1

    The Capacity of a Single Instance: Little's Law in Practice

    Every service can be modelled as a queuing system. Little's Law connects throughput, latency, and concurrency – and reveals where the bottleneck is.
  3. Part 2

    Kingman's formula: phantom traffic jams in the queue

    A correctly sized service degrades long before it reaches its capacity limit. The cause is variability – and Kingman's formula explains why, as the load increases, there is a multiplication of response times.
  4. Part 3

    Amdahl's Law and the Limits to Growth

    Despite having more threads, cores and money, throughput still plateaus. Almost every system has a component that cannot be parallelised. Amdahl's Law demonstrates how this limits scalability and explains why investing in better architecture is often more effective than investing in more hardware.