New Tool Accepted

Thursday, December 12, 2019 21:39

LibReDE: a library for resource demand estimation

A new tool was accepted for SPEC RG's repository of peer-reviewed tools for quantitative system evaluation and analysis: LibReDE.

LibReDE is a library for resource demand estimation. Resource demands are a common input parameter to stochastic performance models (e.g., Queueing Networks, or Queueing Petri Nets). LibReDE helps to determine resource demand values based on monitoring data from a system (e.g., CPU utilization, response time, or throughput).

[A resource demand is the time a unit of work (e.g., request or transaction) spends obtaining service from a resource (e.g., CPU or hard disk) in a system. Resource demands are input parameters of widely used stochastic performance formalisms (e.g., Queueing Networks or Queueing Petri Nets). In order to obtain accurate performance predictions for a system, a performance engineer needs to determine representative values for the resource demands during performance model construction.]{lang=“EN-US”}

[Given that there are no publicly available implementations of estimation approaches, a performance engineer is currently forced to implement estimation approaches on his own. This is a time-consuming and error-prone task. LibReDE is a library supporting performance engineers to determine resource demands by providing a set of ready-to-use implementations of estimation approaches. Based on the actual system and the available monitoring data, the estimation library can automatically determine a set of candidate estimation approaches and execute them. A performance engineer can then validate the resulting resource demand estimates and select the approach that yields the best results. Furthermore, the library also provides a framework that can be used as a basis by developers of estimation approaches. Through reuse, the effort for adapting existing estimation approaches or for implementing new ones, can be significantly reduced.]{lang=“EN-US”}</tools/>.