Benchmarks and Executions
In Theodolite, we distinguish between the static description of a scalability benchmark and its execution.
Benchmarks define what should be executed in scalability experiments. They consist of a system under test (SUT) and an associated load generator, where both SUT and load generator are represented as sets of Kubernetes resources such as Pods, Services, or ConfigMaps. Additionally, benchmarks define one or more load and resource types, scalability can be evaluated for.
Benchmarks are portable, that is they can be provided by standardization organizations, researchers, etc. and installed by other researchers, engineers, etc. They do not have a life-cycle. Instead, to run a benchmark, an execution of a benchmark is created and passed to Theodolite.
An execution represents a one-time execution of a benchmark with a specific configuration. Hence, an execution refers to a benchmark and one of the load and resource types defined by that benchmark. Additionally, executions allow to alter the SUT or the load generator to evaluate regarding specific configurations.
Executions also describe details regarding the scalability measurement method, such as for long experiment should be performed or how often they should be repeated.
In contrast to benchmarks, execution have a life-cycle. They can be planned, executed, or aborted. Each execution of a benchmark is represented by an individual entity. This supports repeatability as executions can be archived and shared.
S. Henning and W. Hasselbring. “A Configurable Method for Benchmarking Scalability of Cloud-Native Applications”. In: Empirical Software Engineering 27. 2022. DOI: 10.1007/s10664-022-10162-1.