Created:
1/25/2006 2:25:45 AM

Author:
Przemek Radzikowski

permalink [Permalink]





International Careers & Jobs - An international employment directory, reviewing world-wide top job sites




  • Home  ›
  • Articles  ›
  • Benchmarking Microsoft Virtual Server 2005



| More

Benchmarking Microsoft Virtual Server 2005

Anyone interested in seeing how Microsoft Virtual Server 2005 performs in comparison to a host machine will find this a good read.


 

 

Who should read this?

Anyone interested in or using Microsoft Virtual Server 2005.

Why I decided to run the benchmarks?

I decided to run these benchmarks because there was nothing out on the net regarding the performance of virtual machines relative to the hosting environment.  This is particularly important because at times it's not necessary to know the exact number of instructions per second that a computer executes, rather we would want a ratio of performance degradation that a machine will incur relative to the hosting hardware.

If I asked you whether you could play Quake III on a Pentium IV 2GHz your answer would be yes.  If I asked you if you could still play it on a machine with 20% less grunt, you would reluctantly still say yes but wouldn't be happy playing it.

My point is that a reference performance benchmark set at 100% given to the hosting hardware allows you to make a judgment call on whether the application will perform well on a system with n% less processing power.  Using a percentage throughput of a machine you already know will help you decide whether to put an application onto a virtual machine or physical hardware depending on your requirements.

Benchmark Setup

The Benchmark Software used for this test was : SiSoftware Sandra v2004.10.9.89

The benchmarks were run under a hosting machine running Windows Server 2003 Standard Edition with all relevant security and hot fixes available at time of writing.  Client machine was also running Windows Server 2003 Standard edition with all relevant security and hot fixes.

Hosting Hardware

Interpreting the results

All results have been compiled with ease of comparison in mind the hosting machine has been given a score of 100%, meaning the fastest possible operation under ideal conditions.  The results of the hosted or client machine will always be slower by nature of design and will be given a percentage score out of the theoretical maximum attained by the host hardware (reference hardware).

Benchmark Results

Benchmark Type Results Performance relative to Host
CPU Arithmetic

89%

91%

CPU Multi-Media

89%

85%

File System

72%

Network / LAN Bandwidth

90%

Memory Bandwidth

44%

43%

Cache and Memory



Summary

In general the virtual machine fared quite well.  Surprisingly network performance was above that which I expected, especially after real production environment tests proved otherwise.   The two largest performance hits are Memory Bandwidth and File System coming in at 44% and 72% respectively. 

Memory Bandwidth

Memory performance will have a huge impact on applications that require a lot of memory paging or general computation requiring high memory I/O.

File System

Any applications that require lots of Disk I/O will see a significant degradation in performance.  Also if your virtual machine is running low on RAM and Virtual Memory paging is required you will have a compound effect on the performance of the machine. 

The first hit will come in the form of paging current memory to disk (at 72% capacity), never good.  The moment you use your hard disk as temporary cache or virtual memory the HDD becomes the slowest component in your machine.  The second performance hit comes from the virtual machine .vhd file (virtual hard disk) itself.  The fact that the .vhd is sitting on top of the hosting file system it will be subject to all file I/O required to operate the host machine.  If you setup the .vhd as a dynamically expanding volume the file will grow when ever it needs the space (up to the maximum specified size). 

Imagine now the situation that you have 10 such virtual machines running on a host.  Each and every one of the machines will be using some virtual memory committed to disk and each will gradually expand the volume as needed.  Eventually you will end up with massive fragmentation all over your volume.

I will be conducting further benchmarks to compare VMWare ESX 2.5, GSX and Virtual Server 2005 Enterprise side by side.  The benchmarks will take into account CPU configurations and Hyper-threding.  Stay tuned.

Further Reading

Read our other review: Benchmarking VMware ESX Server 2.5 vs Microsoft Virtual Server 2005 Enterprise Edition

 

permalink [Permalink] - Updated: Monday, October 28, 2013





| More

 

Articles of Interest


International Careers & Jobs - An international employment directory, reviewing world-wide top job sites


 
 
(c) Capitalhead Pty Ltd
Contact Capitalhead About Us Articles & Publications Partners Solutions & Services Products Valid XHTML Valid CSS