How performance is measured?
Veröffentlicht Februar, 2026 | 5 Min reading time
For users of an application there is only one relevant performance-metric:
“How long does it take the system to respond to my input?”
This, however, does not tell us anything about what’s going on inside the system.
The relationship between user and performance engineer is the same as a driver and a mechanic. The driver measures the performance of the car by its ability to get them where they want to, the mechanic however needs to know what’s going on inside, if he wants to maintain the car’s ability to do so.
To this end four central Key Performance Indicators (KPIs) have been established (while there are more that can also be used, these four should never be ignored):
- Availability
- Response Time
- Throughput
- Utilization
Together these metrics describe whether a system is accessible, how quickly it responds, how much it can process, and how heavily it is loaded in the process. They also give us the ability to systematically understand, compare, and improve the system.
Availability: Is the System even accessible?
Availability describes whether a system is accessible and functional. By definition, an unavailable system has no performance. No matter how fast it would be under ideal conditions.
This does not only refer to complete outages. If parts of a system do not respond they are also not available.
Availability is therefore the foundation of every further performance analysis.
Response Time: How long do the users wait?
Response time measures the time between a request and the corresponding response. It is the KPI that users perceive most directly.
Response time is rarely constant. It varies depending on load, data, volume, or system status. These fluctuations can be very meaningful and should be recorded together with the average value.
Throughput: How much load can the system stem?
Throughput describes how many requests a system processes in a given period of time, for example: requests per second.
This KPI is particularly relevant when many users access a system at the same time, or when processes run in parallel. A system can have good response times as long as there is little traffic. Nevertheless, it can quickly reach its limits under higher loads.
Utilization: How heavily are resources being used?
Utilization measures how heavily the system’s available resources are being used. These include CPU, memory, and network, among others. High utilization is not bad per se. It becomes critical when resources are pushed beyond their limits.
Each of these KPIs highlights a specific aspect of the system. Together they provide a good approximation of performance.