Think of a transportation system: one fast motorcycle may reach its destination quickly, while a bus may move far more people in the same hour. Computing performance works the same way: it measures how efficiently or quickly a system completes work, using different measures for different goals. section 1.2
Two common ways to measure performance are:
- Latency measures how long one task takes from start to finish. Faster completion means lower latency. section 1.2
- Throughput measures how many tasks a system handles during a period of time. A system can have higher throughput even when each individual task is slower. section 1.2
As processor clock speeds stopped rising because of physical limits, manufacturers added multiple processing resources instead. Concurrency and parallel hardware therefore aim to improve performance by using those resources efficiently. Choosing effective operations requires understanding the capabilities of the computer system executing the algorithm, even without focusing on a particular programming language. section 1.2 section 3.0
