IT Innovation Centre will present the latest results of benchmarking the BonFIRE platform and a new performance prediction technique based on algorithmic patterns.
Title: Benchmarking and Performance Prediction of Cloud Applications
Authors: Michael J. Boniface, Vegard Engen, Juri Papay and Stephen C. Phillips
The problems that we have addressed are motivated by assisting Software-as-a-Service (SaaS) providers in the decision making about the selection of Infrastructure-as-a-Service (IaaS) providers and resources required for hosting applications in the Cloud with guarantees on Quality of Service (QoS). In practice these decisions are often ad hoc, most likely requiring users to run their application on resources of different IaaS providers. This can be a time consuming and costly exercise, which is worsened by a lack of transferability since the descriptions of Cloud resources are often ambiguous and can vary between different IaaS providers. The need for predicting the QoS of Cloud applications has driven the development of performance modelling tools within the Platform-as-a-Service (PaaS) layer. We have empirically investigated application-focused benchmarks for characterising compute resources of various Cloud providers and used the collected data for performance modelling of applications.