-4

Our organization is load testing our application using virtual servers via Hyper-V to see what the user load can be using fair equipment on a single box setup.

The developer group questioned the validity of the tests given the normal use of the box by the other virtual machines. IT admins answered that it is an acceptable platform to load test on because it has its own CPUs, memory and disks allocated.

Is their answer mostly correct? What hardware factors may be considered bottle necks given the other virtual machines when testing our application? For example, would bus speed be a concern or network IO?

The application consists of a windows service written using the 4.0 .NET Framework and SQL Server 2008 R2.

Please let me try to clarify the question better. If an application on a virtual machine is being load-tested, how much would the results be affected by other virtual machines on the same host/server/hardware? If the virtual machine allocated resources xyz, how would a non-virtual machine with hardware xyz compare?

sean
  • 101
  • 1
  • 3
    Your question is really unclear; perhaps you could clean it up a bit. I think you're saying that you have a Hyper-V Host, with a VM running your application. The Application is being load tested in this environment and the developers are questioning the validity of the testing (possibly because of the effects of running multiple VMs on the same host, but I'm not sure). There are no typical bottle necks specific to Hyper-V. Virtualized environments commonly suffer a small penalty due to their nature, but it's mostly across the board. It also depends on the hardware used. – Chris S Apr 02 '12 at 22:28
  • Are you trying to compare this to other hypervisors? – tony roth Apr 02 '12 at 22:46
  • Thank you Chris, you are speaking to the question I have. I tried to phrase it better. – sean Apr 03 '12 at 14:59

1 Answers1

1

Sean, the answer you are looking for is far too-wide.

Anything can effect the load testing:

Hardware - speed of hdd/sas, raid config, speed of ram, processor, no. of cores, network adapter speed/config, bus speed etc

Software - sql db setup, the way application pools configured..

Network - type of connections (1GB or 100MB or even load balanced/ aggregated config?), jumbo frames enabled?, network adapter TOE? generally I/O on the network, multicast, broadcast etc

When you want to see the strain of a particular system based on number of user load, you should always try and compare it with another system. This will give you a better understanding and can eliminate the common uncontrollable issues such as the switch I/O on a live network.

Cold T
  • 2,401
  • 2
  • 17
  • 29