Server Virtualization: Definitions, Pros and Cons

With rising costs, budget constraints, increasing costs for energy, floor space, etc. virtualization for servers is making a big push. It has been around for a while first with servers for Internet web sites but now it is making the move into the enterprise, allowing companies to do more with less hardware. But is the push to virtualization all about money (not that money alone would not be a major driving force) and what is the current state of the art with virtualization?

Definition

Server virtualization is defined as the pseudo-packaging of server resources from end users. This includes the number and identity of individual physical servers, processors, and operating systems. The IT department (sorry, had to mention them) uses a software application that allows them to divide one physical server, usually beefed up with fast and multiple processors, lots of disk space and boatloads of RAM into multiple isolated virtual environments. These virtual environments, often called virtual private servers, hold space on the system in files that emulate a real piece of hardware running whatever it is configured to run.

Approaches to Virtualization
There are three basic ways you can perform virtualization: one at the operating system level, one being the paravirtual model (PVM) and the most popular, the virtual machine model.

NOTE: When referencing guests in all three models, we are referring to the ‘Virtual Operating System’ that is running.

The Virtual Machine model uses the host/guest paradigm. Each guest or virtual machine runs on and in a virtual imitation of the hardware layer. Software allows configuration of these virtual machines with respect to how much disk will be dedicated, how many processors, how much RAM and even what flavor of operating system from which vendor will be used. By using this approach, the guest operating system can run without modifications. Another advantage is that it allows administrators to create guests that use different operating systems. The guest operating system has no knowledge of the host's operating system because it is not aware that it's not running on real hardware. This is handled by the virtual server application software that hosts the virtual server. VMware and Microsoft Virtual Server both use the virtual machine model and handle all access to the real hardware for these guests via a Hypervisor or virtual machine manager. The Hypervisor is a program that allows multiple operating systems to share a single hardware host, allowing each guest operating system to have access to the host's processor, memory, and other resources all to itself. In reality, the hypervisor is actually controlling the host processor and resources, allocating what is needed to each operating system in turn and making sure that the guest operating systems (called virtual machines) cannot disrupt each other.

The Paravirtual Machine model (PVM) is also based on the host/guest paradigm just as the VMM did. And just as the VMM did, it uses a virtual machine monitor too. The main difference is that in the paravirtual machine model the virtual machine monitor (VMM) actually modifies the guest operating system's code. This modification process is called porting. Porting allows the VMM to utilize privileged systems calls sparingly and thus makes the virtualization run more efficiently. Just as with virtual machines, the paravirtual machines are capable of running multiple operating systems. Xen and UML (User Mode Linux) both use the paravirtual machine model to accomplish virtualization.

The third method, virtualization at the OS level, works a little differently. OS Level virtualization isn't based on the host/guest paradigm. The host runs as a single OS kernel as its core in the OS level model. That OS kernel exports operating system functionality to each of the guests. A limitation of this implementation is that the "guests" must use the same operating system as the host. Even though they are all running the same operating system, the distributed architecture eliminates system calls between layers, which reduce CPU usage overhead. OS level virtualization also requires that each partition remain strictly isolated from its neighbors so that a failure or security breach in one partition isn't able to affect any of the other partitions. Examples of this type of implementation are Virtuozzo and Solaris Zones through opensolaris which both use OS-level virtualization.

More Bang for the Buck
Server virtualization may not be for everyone. If you have very high-end applications that consume lots of resources and do lots of disk and memory access, you may not want those server applications placed in a virtual environment. Lesser-used or departmental server functions can be carried out nicely in a virtual environment. This will allow the purchase of a server loaded with disk and RAM, take only one place in your rack of servers and carry out multiple server functionality. In addition to the things already mentioned, server virtualization can be used to more efficiently use server resources, improve server availability, assist in disaster recovery, testing and development and to centralize server administration. It is well worth looking at this as an option when you are looking at purchasing new hardware to replace aging servers. Just make sure that what you want to run will work in the virtual environment of your choice. Check with the manufacturer of the software you intend to run in that virtual environment and make sure they support it and they license the software for use in a virtual environment. Not having the vendor’s blessing can be a deal breaker when it comes to support!

How can I get more information on this subject?

We strive to stay up to date and provide you with the latest information available on technical subjects that are gaining momentum. Technical trade publications are a good place to start, but getting information from a partner that uses and supports technology is a great way to stay on the leading edge and away from the bleeding edge of technology. If you have questions or comments about this article or if there is anything new on the subject, contact me. I would be happy to respond!

All product names / logos, company names / logos are copyrights of their respective holders. John Boline is an MCSE, CNE, USE and a member of the Network Professional Association. The content herein is often based on late-breaking events. Much of the material is based on information from sources that are believed to be reliable. Hagerman & Company, Inc. disclaims all warranties as to the ultimate accuracy or completeness of the information. Hagerman & Company, Inc. and its employees shall have no liability for errors, omissions or inadequacies in the information contained within this article or for any interpretations thereof. The recommendations, positions and best practice policies outlined herein represent Hagerman & Company, Inc. initial analysis and therefore are subject to change as further information which may have bearing on these positions is made available. The reader assumes sole responsibility for the selection of these materials to achieve its intended results. The opinions expressed herein are subject to change without notice. Hagerman & Company, Inc. assumes no obligation to update the forward-looking statements made in this newsletter to reflect any change in circumstances, after the date of publication. Entire contents © 2009 Hagerman & Company, Inc. All rights reserved. Reproduction of this publication in any form without prior written permission is forbidden