Amdahl’s Law

As explained in Learn Concurrent Programming with Go

A performance model stating that the overall speedup from optimizing or parallelizing a system is limited by the fraction of execution that remains sequential.

Amdahl’s law states that optimizing or parallelizing one part of a system cannot overcome the time spent in sequential work. The non-parallel portion becomes the bottleneck limiting overall speedup. section 1.6.1

Building a house with more workers

A house project may be 95% parallelizable, while tasks such as moving materials along a single road remain sequential. Even with 512 builders, the project finishes only about 19 times faster; adding more workers eventually wastes money. section 1.6.1

Use Amdahl’s law to reason about scalability:

• Find the work that must happen one step at a time. section 1.6.1
• Estimate whether additional processors can improve the parallel portion enough to matter. section 1.6.1
• Avoid assuming that more workers or processors will produce proportional gains. section 1.6.1

Amdahl’s law describes fixed-size problems. Gustafson’s law offers a complementary view: increasing the problem size can keep additional resources busy when the serial portion stays relatively constant. section 1.6.2 section 1.7

Figure 1.6 The speedup against the number of processors according to Amdahl’s law

Shows how theoretical speedup changes as processor count increases for different parallel-to-sequential ratios. section 1.6.1

How Learn Concurrent Programming with Go covers it

How deep it goes

Defines sequential bottlenecks and illustrates speedup limits with construction scenarios and a processor-scaling figure, without deriving the formula.

Examples

Uses house construction, a narrow transport route, processor-count charts, and a 95%-parallel workload reaching roughly 19-fold speedup.

What it assumes

Assumes only a distinction between sequential and parallel work, which it develops through a construction analogy.

What you can do after

Evaluate scalability limits and choose whether additional parallel resources are likely to improve a fixed-size workload.

How it connects

Connects sequential execution, parallelism, scalability, processor resources, and Gustafson’s law.

liveBookmanning.com

Read Learn Concurrent Programming with Go on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Grokking ConcurrencyAmdahl's law is a formula for estimating the maximum speedup of a computation from its sequential fraction and available processing resources.
• 

  liveBookmanning.com
  LatencyAmdahl's law is a performance model that estimates the maximum speedup from parallelizing a fixed-size computation according to its parallelizable fraction and number of execution units.

Compare how 3 books explain Amdahl’s Law →

Related concepts

• speedupmeasures · Speedup is the ratio of a baseline execution time to an alternative execution time for the same computation.
• parallelismuses · Parallelism is the concurrent execution of multiple computational operations, often across multiple processing resources.
• meanuses · A mean is a measure of central tendency, calculated as the arithmetic average of values or, for a random variable, as the probability-weighted average of its possible values.

This explainer was generated by Manning's knowledge graph from the text of Learn Concurrent Programming with Go. The section links point to the parts of the book it draws on.