Comparing performance of public/private clouds
The Emulation-as-a-Service (EaaS) framework (https://gitlab.com/emulation-as-a-service/eaas-server) lets users access preserved digital assets using their Web browser. To do so, it starts a software emulator (e.g., QEMU) per emulation session in a virtual machine instance in the cloud (multiple emulation sessions can run on the same virtual machine). The virtual disk images which are connected to the emulators are stored in S3-compatible storage buckets and are accessed via HTTP(S) range requests while the emulator is running.
This nested approach (emulator on VM) presents a computationally relatively challenging scenario for computing clouds, which different cloud providers handle differently well. However, a benchmark to judge the performance (e.g., latency, starting delay, availability, interaction between computing cloud and S3 storage, ... vs. price) of different installations on different clouds is currently missing in the EaaS project.
The goal of this bachelor/master project/thesis is to examine how well the OpenStack-based private cloud operated by the University of Freiburg as part of the bwCloud project performs with this scenario in comparison to different (commercial) public cloud providers (e.g., Google Compute Engine, OVH, Vultr, possibly together with external S3-compatible storage providers like Backblaze). As part of the project/thesis, a benchmark scenario should be developed and tested on different clouds.
Kontakt