Wednesday 29 August 2012

vSphere vs Hyper-V performance showdown


The inspiration for this came about from collecting data in both of these environments. I had athene® bringing in data from both vSphere and Hyper-V and thought it would be interesting to compare these two platforms by running some workloads against them.

For those of you who use these virtualized environments, you know there are many metrics that you can collect from these two environments.

All the metrics that I have collected are the standard types of metrics that you would use for on-going capacity management and were collected from the hypervisor.

Architecture

I’ll start by taking a look at the architecture:

vSphere for ESXi version 5
 
VMware advertises this as having a small hypervisor footprint, and it does.
Hyper-V
 
Microsoft is quite a bit different in the way that it is constructed, when you install the Hyper-V role with Windows 2008 R2 server, it creates what Microsoft terminology refers to as a partition.

Microsoft refers to two different types of partitions:
Enlightened – has an integration services pack installed which provides extra features.
Unenlightened –a virtual machine that is running on its own with no special features.
The equivalent of the integration packs for VMware is VMware tools.
Having these integration packs or VMware tools installed is highly desirable, unless there is some particular reason that you should run without them, as they help you to realize the full benefits of virtualization.
 
Windows Hyper-V runs with a larger footprint.
 
On Friday I'll be looking at the memory management features of both and later in this series will examine how they perform under some tests that I ran, so keep following.
 
Dale Feiste
Consultant
 

 
 

 

2 comments:

  1. Hi Dale, my company is looking at doing similar benchmarks on vsphere5 and hyper-v3. Can you please advise which tools were used to perform the benchmarks?

    Thanks for your help.

    ReplyDelete
    Replies
    1. Hi

      Thanks for your question. I generated a standard workload using simple perl scripts to ensure each system ran identical loads. I then analyzed the data using Metron's athene®.

      Dale

      Delete