Posted by admin on Sep 29, 2009 in Performance, VMware | View Comments
Myself along with many other admins supporting Virtual Infrastructures are always looking to increase their ability to diagnose performance related issues. As new software and hardware technology is released we get more tools to help and more things to think about.
To help with the issue VMware has published a few helpful documents. I have broken down the different documents and included links to the originals at VMware.
Hyper-Threading is a common question about how ESX handles this technology. I’m going to assume that if you are reading this that you know what HT is. Hyper-Threading has been supported in ESX since version 2, since ESX is aware of HT it treats logical cores different than physical cores. Using HT does offload some of the SMP calculations from the OS and allows the processor hardware to do the work at a faster rate. You can read the whole documente here.
The next VMware article covers the ESX Monitor modes available. It breaks down the different Hardware Virtualization options supported for different ESX features, guest OS and processor families. See full article here.
Next in line is Guest OS based performance. Due to the fact that VMware removes the hardware layer from the VMs touch. This in turn makes Windows perfmon or Top in Unix return results that are not very accurate. So this brings up the question are you using your tradition tools that you would install on your physical servers? If so they are unaware of the benefits being realized by the virtualization software. The tools do not have an accurate view of cpu resources since each VM is getting a fraction of CPU time. See the document here.
The next big item to cover is CPU read time. This value always gets a lot of questions and is confusing to many. To make it as clear as I can Ready Time is the amount of time a Virtual Machine waits to run a process but has not be provided CPU resources on which to execute. There are two ways to get a value for ready time, you can use Esxtop or Virtual Center. If you use Esxtop you will get a percentage value and Virtual Center will give you a time value in milliseconds. To give you an idea 1,000 ms in VC’s report is equal to 5% fom esxtop. Read more on full article here.
read more
Posted by admin on Sep 23, 2009 in VMware, vCenter Server | View Comments
You have probably been hearing some of the buzz around vCenter Chargeback that will be available for VMware vSphere. Its a separate server that will work with vCenter server to all for chargeback of virtual resources. This will help the business to understand the costs of deploying new VMs. The IT side will use this to report and track on resources usage in the Virtual environment.
Within vCenter Chargeback there are several costing models that you will be able to use. First there is the Fixed Costing model that allows you apply a fixed cost for a VM instance. Secondly there is the Allocation Based Costing that offers variable costs per VM based on it’s allocated resources. Lastly you offer Utilization Based Costing that will be based on actual resources utilized.
You will be able to access this data via the vSphere client or a Web browser to allow your business customer the ability to get reports.
Currently Chargeback will support vCenter servers versions 4.0 and Virtual Center 2.5 update 3 and above. This is good news for companies not looking to move to vSphere in the near future.
You will be able to breakdown the cost models into per hour costs that track metrics like CPU, Disk Read/Write, Memory, Network traffice, Storage and a few more. You can choose to also bill for FT, HA and other custom add-ons. You can use cost templates to input values for different types of VMs that might be deployed.
read more
Posted by admin on Sep 22, 2009 in Troubleshooting, VMware | View Comments
If you have ever lost a host to a PSOD (Purple screen of Death) you know that they may as well be written in Greek. Sure you used to think that reading a Windows BSOD was a pain, but the VMware PSOD takes this to a new level.
In the past trying to break down what caused your host to crash on your own was a big task. Most folks would require a call to VMware support if the host did not come back up afterwards. But now VMware has since written a nice document on breaking down the many sections of the PSOD screen. This will give you some insight on what might be going on with your ESX host.
You can have a look at the VMware document here.
read more
Posted by admin on Sep 17, 2009 in VMware, vCloud | View Comments
I was luck enough to be invited for a usability study on VMware vCloud. I had to sign a NDA so I cannot say anything about the product. But we spent about 90 minutes going through screen shots of the product. We talked about how I would use the product, what I thought might be the use of screens shown to me.
Overall the vCloud product is not ready for release yet and who knows when it will be. But it really kicked butt. It was packed full of features and seemed pretty flexible. I work with corporations that would love to have this today for Internal Cloud usage.
The word vCloud has been coming up in discussions for atleast a year now. It’s good to see that its not just vaporware and VMware seems to have something working up there.
read more
Posted by admin on Sep 15, 2009 in VMware | View Comments
Today VMware dropped a whitepaper on the big bump in performance of the VMXNET3 virtual network device used in vSphere. It overs a significant performance increase over the previous version under several configurations and protocols. Along with the performance gains it reduces overhead on the CPU usuage.
Some highlights of this paper are:
(1) Throughput gains of up to 92% for 10G TCP/IPv4 Rx workloads with large socket buffer, which greatly improves bulk data transfer performance in a data center environment.
(2) Dramatic gains across all configurations of IPv6 traffic, with significant CPU usage reduction and throughput improvement over enhanced VMXNET2.
You can see the new relases here and download the whitepaper here.
read more
