FIFE: an Infrastructure-as-Code Based Framework for Evaluating VM Instances from Multiple Clouds
Yuhui Lin, Jack Briggs, Adam Barker · 2020
To choose an optimal VM, Cloud users often need to step a process of evaluating the performance of VMs by benchmarking or running a black-box search technique such as Bayesian optimisation. To facilitate the process, we develop a generic and highly configurable Framework with Infrastructure-as-Code (IaC) support For VM Evaluation (FIFE). FIFE abstract the process as a searcher, selector, deployer and interpreter. It allows users to specify the target VM sets and evaluation objectives with JSON to automate the process. We demonstrate the use of the framework by setting up of a Bayesian optimization VM searching system. We evaluate the system with various experimental setups, i.e. different combinations of cloud provider numbers and parallel search. The results show that the search efficiency remains the same for the case when the search space is consist of VM from multiple cloud providers, and the parallel search can significantly reduce search time when the number of parallelisation is set properly.