Microsoft may have been late to the server virtualisation party, but it’s Hyper-V beginning to see success up against the more established VMware and Xen hypervisors.
Adelaide-based Windows hosting company Expressnet first dabbled with virtualisation with Virtuozzo, but, according to managing director Simon Holman, the stability and performance was inadequate.
"Software services were not restarting when we restarted the server," Holman told TechWorld.
Expressnet then investigated VMware and Xen, but "wasn’t overly convinced" at the suitability of those products either.
When Hyper-V came out as part of Windows Server 2008, Holman was impressed with the beta version of it.
"We rolled out a Dell server as a test bed for Hyper-V and we were testing virtualisation from a customer product offering point of view and for our own systems," he said.
The trial was successful and Expressnet went into production and now has customers looking for virtual dedicated servers.
"Customers are happy with performance and its even taken me by surprise," Holman said, adding another server is now being brought online for virtual machines.
The hardware is Dell 2950 servers with two quad-core Xeon processors and 16GB and 32GB of memory, respectively.
The 16GB machine is running 12 VMs and it is anticipated the 32GB machine will run 20 VMs.
Holman said one of the restricting factors of Hyper-V is that it pre-allocates RAM so as soon as the VM is started it allocates the memory to a particular VM and can’t be changed without a re-configuration and reboot.
"Xen and VMware can do that but the performance benefit we are getting outweighs that," he said. "The shut down and boot times are faster in virtual environment and only takes about 40 seconds."
Holman said Hyper-V is lacking in management tools; however, Microsoft has released the its System Centre VM Manager which makes it easier to do advances features like cloning VMs and templating.
"The performance monitoring we did was fairly loose. We ran Xen 4.1 and ESX 3.5 on two boxes with 2GB of RAM and it was more of a gut-feel scenario than proper benchmarking. We found VMware and Xen was the slowest."
Expressnet also leaned towards Hyper-V as a result of legal uncertainty with hosted VMs on VMware.
"In the terms and conditions it was a bit ambiguous whether you can offer the product how we wanted with the base-level ESX server," Holman said. "But that was a secondary concern. We were looking at performance and cost."
Holman also praised the "point-and-click" installation of Hyper-V on Windows Server 2008 data centre edition, which took about 20 minutes. But the virtual network adapters had to be set up manually.
The data centre edition of Server 2008 allows unlimited VMs to be hosted on the physical machine.
The guest operating systems are Windows Server 2008 as it was unclear of Microsoft’s Windows Server 2003 licence allowed migration of a physical system to a virtual one.
Holman has been trying to get a hold of Microsoft’s Linux integration tools for Hyper-V for some compatibility testing, but "they seem exceptionally hard to find".
"Virtualising saves me a lot of time and resources and we’re using less power and servers," he said.
Latest on Linux
- Group pitches Linux for free netbooks from mobile carriers
- OpenSource World offering free admission
- Linux 2.6.30 released, Torvalds proclaims 'new world order'
- Fedora 11 adds virtualization, software management tools
- Android Market to work on Intel's Moblin
- Intel buys Wind River to push Linux
- Intel to acquire Wind River for $884 million
- Acer will use Moblin Linux across its products
- Microsoft to leave smartbooks to Google
- Taiwanese netbook runs both Android and Windows
TechWorld Jobs (beta)
Recent Jobs
TechWorld Blogs
-

TalkingTech
The view from the top of IT with TechWorld Editor Rodney Gedda
-

Entrenched
Cooking up better code, IDG's developers reveal some of their secrets
-

Broadband Voice
Darren Pauli digs in from the front line of Australia's broadband battleground
Recent blog posts
- Nokia remains 'open' to Android amid Symbian renaissance
- KDE's Seigo gives sneak peek at version 4.3
- Was the iPhone 3G S worth queuing up for?
- Has Oracle started its mammoth technology consolidation?
- iPhone 3.0: the detail is the process, not the features
- TechWorld.com.au goes mobile
- Should Dell buy Palm? Stranger things have happened
- A big week for Linux: is user friendliness finally in sight?
- Apple, Android rain on Palm's Pre parade
- The clone attack is becoming unstoppable
Recent comments
- PSP Nintendo
13 hours 14 min ago - Interesting report. You were
1 day 9 hours ago - Are you sure it is in Sydney?
1 day 20 hours ago - The mobile market has
2 days 4 hours ago - Great news.
Sms spam should
3 days 1 hour ago - now what am I gonna do with
3 days 4 hours ago - ozlotteries.com not ozlotto.cm
3 days 5 hours ago - OLAT Release
3 days 15 hours ago - and i was sure i would win...
3 days 20 hours ago - Hi SolidRadicle,
I am looking
3 days 20 hours ago - Not if I can help it
3 days 20 hours ago - Ozlotto Tips Scam
4 days 1 hour ago - Great post.
It's very
4 days 1 hour ago - Excellent review! I'm glad
5 days 22 hours ago - iTunes Helper
1 week 1 day ago - Update the link to OrangeHRM web site
1 week 2 days ago - Very informative article
1 week 2 days ago - Google Chrome is still being directed to bing instead of google
1 week 2 days ago - regd: Software Magazine
1 week 2 days ago - I seem to have missed a point
1 week 3 days ago










Comments
Clarifaction of article
There seems to be some confusion by the three people who have posted comments over the testing procedure that was used when testing the virtualisation platforms.
All three platforms were tested on the same host server which was running 8 GB of RAM with two virtual machines each with 2 GB of RAM. One running SQL 2005 and the other running Windows 2008 Standard with IIS 7.0. So the evaluation platform was standard across VMWare, Xen and Hyper-V. This ensured that both running VM's had the same resources and that the host server had ample resources in reserve. This may not have been explained properly in the article.
If anyone has any concerns over the testing procedure or the results then I invite them to contact me directly via the Expressnet website at http://www.expressnet.com.au to discuss further.
Regards
Simon Holman
Managing Director
Expressnet
How About the Correct Title?
First, change the title. VMware and Xen weren't kicked out - they were never in there. VIrtuozzo (not even in the tile by the way) was the one that was kicked out.
Second, you tried to do an eval with a box with 2 GB of RAM? For real? I know you have money since you're running a hosting company. And you're surprised that Microsoft did better when it was running on a system with 16 GB of RAM?
Please, if you aren't even serious about a product then why even test it? Just take your pretty Microsoft check, cash it, and be done with it. And for the person writing the article - shame on you. You really need to try harder in the future. At least we get a good example of how to write a bad article for journalism class next time.
who are you kidding?
For Hyper-V
>The 16GB machine is running 12 VMs and it is anticipated the 32GB machine will run 20 VMs.
and VMware/Xen
>The performance monitoring we did was fairly loose. We ran Xen 4.1 and ESX 3.5 on two boxes with >2GB of RAM
Gimme a break! Tell me, what did your Hyper-V testbed look like? Does Hyper-V _even_ boot on a box with 2GB of memory? How much RAM is given to the Hyper-V VMs out of the 16 or 32 GB, and how much RAM was given to the Xen/VMware vms from the _2_ GB ?
Yes, your performance <cough>monitoring</cough> is absolutely loose.
Seriously, how much did they pay for this advertisement
Are you kidding me, did you actually write a review of VMware vs MSFT based off a 100% MSFT house? So the director of IT at any company is going to a production environment of a basically beta version vs. that of a technology that has been built upon for 10 years. Who needs management capabilities, who needs a business continuity/disaster recovery plan when you have such a stable solution in hyper-v? I'll stick with VMW and keep my data center up and running...I'll also keep my job, sleep at night and know that my customers will be in good hands.
Post new comment