David - your assumption is incorrect. what a Virtual Machine allows you to do is run multiple "computers" simultaneously on the same hardware. From your description, it sounds like you could *replace* all the other computers around the house by running VMs on your main 24 core server. You could easily allocate 3 cores for one Virtual Machine, 1 core for another, and so on - the only limitation would be memory and the performance you are willing to tolerate either on the main system or inside the Virtual Machines.

Any other kind of sharing amongst disparate machines would be done via network protocols (CIFS or NFS for filespace). There is a commercial product, ScaleMP, that supposedly allows you to aggregate memory from several different machines into one larger configuration, but again - that is not what you are describing.

Yes. The basic concept of virtualization means that the 4 VMs you configured will be multiplexed (share) on a single hardware. If you use 4 separate physical machines for each of them, there will be no sharing and contention of resources. The performance will be definitely better in the latter scenario. However, virtualization will increase resource utilization.

Thanks a lot Mr. Shuja

I would like to repeat my question as:

I am using 4 individual machines (hardware not shared) and each machine is running window 7. I am running Ubuntu on VMWare on each window 7 separately. Since I have 4 machines I could run Ubuntu on native hardware instead of VMWare. But I did not. So now what will be the difference?

Sudhakar, since you have chosen to run Ubuntu in a VM, understand that you will have a performance penalty of Windows 7 processing + VMWare processing + Ubuntu processing - before you actually run any application under Ubuntu. As you add layers of software between the physical hardware and the application you choose to run, there will be a performance hit at each layer. Unless you have a dedicated Ubuntu machine, you won't be able to tell how big a performance hit you are suffering.

Yes.

The following article has an interview with a senior architect of VMWare

gives some details for modern processors which have hypervisor support

The architect speaks of performance hits ranging from

1-30% depending on the percentage of time in the kernel and in I/O .

http://www.anandtech.com/show/3827/virtualization-ask-the-experts-1

The big reasons for virtualization come from

1) merging several VMs each running low cpu processes

onto a single compute blade to save on hardware costs

2) migrating VM from hardware that is exhibiting problems to other machines without interruption of service

3) the ability to start up a new VM from a set of preconfigured VMs very quickly

4) the ability to suspend a VM and start it back up again. (This cycle soaker for off-hours computational use where the process is checkpointed on shutdown because the entire VM machine state is saved.)

I tholught that virtualization allowed for one machine to use the unused resources of all the other machines on the local network, and if there was a powerful server on the network, it would be feasible to use simple, inexpensive physical machines that have the power of the server. (I just happen to have a server with 4 processors, each with 6 cores, and about 5 Tb of Hard Drive on my home network. I am planning to virtualize so that all the other physical machines scattered around the house can use that power.) Please tell me that I am not wrong about what virtualiation will do for me.

David - your assumption is incorrect. what a Virtual Machine allows you to do is run multiple "computers" simultaneously on the same hardware. From your description, it sounds like you could *replace* all the other computers around the house by running VMs on your main 24 core server. You could easily allocate 3 cores for one Virtual Machine, 1 core for another, and so on - the only limitation would be memory and the performance you are willing to tolerate either on the main system or inside the Virtual Machines.

Any other kind of sharing amongst disparate machines would be done via network protocols (CIFS or NFS for filespace). There is a commercial product, ScaleMP, that supposedly allows you to aggregate memory from several different machines into one larger configuration, but again - that is not what you are describing.

Quick answer: every time you virtualize something, you get penalty over running it on the bare hardware. That means that the OS running in virtual machine will run slower than if it would run directly on that same machine without virtualization layer.

That's true for any OS on commodity hardware, with exception for some exotic hardware that is purposefully built to support virtualization from the ground up - and that's not the off-the-shelf computer.

Full virtualization like using KVM should have obvious performance difference. At least my tests on disk I/O show this.

Paravirtualization like Xen, may have less performance penalty.

Regards.

I have used VMWare and Virtual Box only. Please let me know more about KVM and Xen. Thanks in advance.

KVM (kernel-based VM) and Xen are both open source, widely used in the academic community. KVM is a built-in module of Linux. Just check the manual on how to use it, it's direct. Xen needs to be installed, involves more procedures to be used.

Sudhakar,

Definitely Yes to your question. We had one experiment Me and my mate established a mini cloud and created Hadoop Cluster with 4 nodes. We also had a off-the shelf commodity machines. Using those machines we created a Hadoop Cluster. We then did some experiments in both environments(one on Cloud and another on Commodity machines). We found that the performance is somewhat similar. But if you talk about flexibility and scalability, we found Cluster on Cloud more appropriate. Virtualization is far better than Physical ones. in my opinion.

You have an overhead to manage the complexity of virtual machines. More modern processors have instructions to deal with same virtualized environements.

If you have a lot of VPS's in a server you will feel some difference but with a few the level of resources will be more or less the same.

VPS's issues will appar more on disk sharing and usually a machine with more load will benefit from having more memory. Virtualized environment make it easier to have a bigger percentage of resources being used.

At this moment i have same nodes with 2x6 cores to provide web hosting services. Xen and KVM makes a bigger isolation and openvz allows to have access to all files from root node.

Definitely, virtualization has overheads. In some scenarios, overheads are huge. In general, CPU and DRAM virtualization overheads are relatively slight due to the adoption of hardware assisted virtualization technologies such as Intel-VT. However, the disk virtualization or emulation has considerable overheads, because the virtual disk emulation is implemented in user space. Especially, if the storage media is SSD, the I/O latency overhead of virtual disk becomes obvious. For example, QEMU emulation adds an additional 60-70 micro seconds to the I/O path, the bare metal SSD I/O latency is about 40 micro seconds, so virtualization doubles or triples the storage I/O latency compared with the native access.        

One of my evaluation paper discussed the overheads of storage virtualization. If you don't mind, you can access it to get some insights. Thanks. 

PS: The evaluation paper is titled "Understanding the Impact of Cache Locations on Storage Performance and Energy Consumption of Virtualization Systems"