Archive for November, 2008
Last week I attended a meeting with several guys from Citrix. Purpose of the meeting was to convince us that the latest releases of Citrix XenServer are mature products for the enterprise market. Why this meeting? Because lot’s is said about Xen (the non commercial product) and XenServer (Citrix commercial product), but for now it only has a 3% marketshare. So when customers ask us for a reference it far more difficult to give a reference than for example VmWare. But back to the point, What did Citrix tell us to convince us?
I stumbled upon a nice white paper on the VMware blog concerning virtualizing SQL servers. The conclusion of the article (how surprising) is that in high workload scenario’s it can be really interesting and usefull to virtualize your Microsoft SQL database servers.
I really advise you to read the full article, because every situation is different but for those who are lazy:
Based on the test data presented in this paper, we can conclude that:
- As the number of virtual machines is increased ESX offers linear scaling of performance until the physical CPUs are saturated.
- Fairness in resource sharing makes ESX a robust platform for hosting and consolidating virtual machines running SQL Server workloads.
- While resources are available, the CPU load on the host has minimal impact on the performance of applications running in virtual machines on that host.
- Virtual machines get the resources they need when resources are not overcommitted. Resources not used by idle virtual machines are dynamically allocated to other virtual machines that need resources without affecting performance.
You can read the entire article here on the VMware Blog (VROOM section)
I was configuring my ESX whitebox (ESXi 3.5 update 3) and I was having some troubles with installing new VM’s. I was having problems during the installation of Windows 2003 from a mounted ISO. Some files couldn’t be copied during installation:

After I retried to copy the files, it sometimes continued copying, but after a while, I got a nice bluescreen:
From the error’s it is not directly clear that it is a memory problem, but the first thing I tried was to pull out some memory modules. The problems disappeared immediately.
Now this is why VMware advises to test the memory for 72 hours ( you can use Memtest86) before you put an ESX server in production.
A lot of dutch bloggers already mentioned that the Dutch VMUG event is on December 12th and that there is still room for a couple of attendees (600 max). So if you haven’t already registered, please do now, it is well worth it.
The agenda looks pretty good with sessions about VDI, Powershell, storage, disaster recovery and a lot more. Besides the agenda, this event is also about networking and discussion on virtualization. Expect a coverage on virtualfuture.info. We will also try to capture some video-footage.
Bloggers on the scene:
- Aleks: virtualistic.nl
- Arne: ict-freak.nl
- Bouke: jume.nl
- Duncan: Yellow-Bricks.com
- Eric: NTPRO.NL
- Gabrie: GabesVirtualWorld.com
- Joep: virtuallifestyle.nl
- Sven, Matthijs, Johan: virtualfuture.info
Normally when multiple datacenters are created in Virtual Center it is because they located at different sites aka physical datacenters. In this case the administrator created a datacenter that spanned multiple physical locations, which isn’t a good idea.
That’s why he created a new datacenter and cluster at his own location for the hosts at that site, VMotioned all running VM’s to other hosts and set the host in maintenance mode, believing that he could VMotion the running VM’s to the new datacenter. Well it went a bit different than he thought: although Virtual Center wasn’t happy VMotioning the VM’s to the new datacenter (it gave a clear warning that in the new datacenter vSwitches might be missing) it gave an OK, but when the Finish button was clicked it failed all together stating that VMotioning between datacenters isn’t allowed.
Ok this is what he eventually did: he disconnected the host from the original Datacenter\Cluster with all VM’s still running on that host and reconnected that host to the new Datacenter\Cluster. I must admit it was pretty bold to do but it all worked because no ip-addresses/names etc were altered in this process. Whether it is a supported method I really can’t say, the only thing I know it worked for him.
In my opinion it is better to have good design and stick to it because good designs are ment to be flexible, easy to maintain and easy to expand.


