In the previous part of this OpenStack KVM vs Hyper-V compute benchmarking series we talked about how to setup the testing environment and how to run some scenarios. Time for some results!
To begin with, we started with a basic KVM vs Hyper-V scenario:
- Ask Nova to boot a VM
- Wait until Nova reports the VM as active
- Ask Nova to delete the VM
Results become especially interesting when looking for bottlenecks in the infrastructure, so the following results were obtained by running the test with 50 VMs in parallel on a single node with a total of 200 VM.
- Results for KVM running on Xenial Ubuntu 16.04.1 LTS (default kernel version 4.4.0-45-generic):
- Results for Hyper-V / Windows Server 2012 R2:
- Results for Hyper-V / Windows Server 2016:
Each iteration sequence number corresponds to one Rally scenario execution. We can see a few spikes in the graph for Hyper-V Compute with Windows Server 2012 R2. Those are due to a minor bottleneck on Windows Server 2012 R2, which is no longer present in Windows Server 2016.
On average in this scenario, KVM is ~32% slower than Hyper-V on Windows Server 2016, but ~18% faster than Hyper-V on Windows Server 2012 R2.
Not only is Hyper-V on Windows Server 2016 faster than Hyper-V on Windows Server 2012 R2, but it’s also significantly faster than KVM!
Ready for more scenarios? part 3 is going to be available soon!