Problem solve Get help with specific problems with your technologies, process and projects.

Hardware for virtualization: How to get more from servers, HBAs and more

Maximize the productivity of virtualization hardware with these tips on servers, network cards, host-bus adaptors and more. Rick Vanover shows how to add more power for less money while still achieving your goal of a successful virtualization implementation.

When planning a virtualized environment, there are a couple of strategies you can employ to make the most of your server investments. Sure, you can overprovision the virtualization hosts as a forward-looking strategy, but if you are working with limited funds there may be some additional factors you may want to consider.

In this tip, I will dig into the main categories of virtualization hardware: servers, network cards, host-bus adaptors and more. I'll show you how you can add more power for less money while still achieving your goal of a successful virtualization implementation.

Server selection
There are many different ways to approach an enterprise virtualization environment from a design perspective. This tip will revolve around the strategy of having the environment composed of a large number of (relatively) smaller servers versus a small number of larger servers with the maximum resources. The thought here is to be able to scale linearly with a larger number of relatively lesser powerful servers so that it is easier for you to add servers from an incremental cost standpoint. Consider the following scenario where you implement a virtualization environment on x86 servers, and you purchase six servers with dual-socket dual-core capabilities. The hardware cost to add additional dual-socket dual-core systems can cost in the range of $4,000 to $9,000 depending on the configuration and brands selected. Whereas a system with a quad-socket quad-core and a maximum memory configuration can start at $10,000 and cost over $30,000 simply in hardware costs for the higher performance model.

The incremental cost of the higher powered servers may prohibit you from expanding in a linear fashion as the virtualization needs increase over time. Of course, there is the school of thought that if you have a very large virtual environment, such as hundreds or thousands of virtualized systems, every host needs to be of the highest available physical resources for the virtual infrastructure.

Networking selection
In the virtualized server environment, you do not want to pin yourself into a corner from the initial provisioning of the host systems. And as the number of virtualized systems increases, the network presence requirements may increase as well. Consider the following example: if you have a small farm of approximately five virtual server hosts you may want to connect guest operating systems to your live server network, your test server network, the client network, and possibly your demilitarized zone (DMZ) network. To avoid bottlenecks of a limited number of interfaces available, consider purchasing the four-port network interfaces available with most servers.

Take for example the Intel PRO/1000 PT Quad Port Server Adapter. This particular card is a good candidate for virtualization implementations with a 2.5 GHz (each direction) bus speed by using the PCI-Express interface. Each of the four ports is capable of running at 1000 Mbps at full duplex. This card has a street price in the $400 range, so it is clearly a great deal for the performance. If you put two of these interfaces in your virtualization host, you can put one (or more) interfaces from each card on each of the networks you wish to place your guest operating systems, thus having redundancy across the two cards. This can put you at four separate networks, with 2 interfaces on each network for your guest operating systems. And if any on-board network interfaces are available from your server, you can configure management interfaces on these ports to keep it isolated from your guest OS traffic.

Storage strategy
For the scenario where you are using a Storage Area Network (SAN) to host your virtual machines there can be some cost saving strategies as well. Unfortunately, the host bus adapter (HBA) products are generally the most expensive accessory that you will add to the server, so make the best out of the investment. One starting point is to have your virtual machine host boot from the SAN. This will eliminate the need for local disks that simply hold the operating system and possibly save you the expense of an array controller (many server models have a default array controller). Further, if needed, you could get higher performance disk and caching on the SAN than you could locally. The hidden benefit here is that you don't have to manage the local discs either as your storage administrator would be doing these functions already for your much larger allocation for your virtual machines.

Another tip is to use dual-port HBA cards. These can reduce the cost and internal expansion port footprint by consolidating interfaces. The HBA cards pose one different scenario in cost, however. Take for example QLogic HBA cards. They are available in single port, dual port, and quad port cards. The cost difference between a single port and a dual port HBA is usually minimal on PCI-Express interfaces, bringing the price per port down. The quad port interfaces; however, have a higher individual list price which is approximately double that of a dual port interface. If you need four HBA ports to your SAN, the better strategy (if you have the available PCI-Express ports) is to use two dual-port HBAs. This is because there is no real price break between a dual port and a quad port – yet you can retain two separate interfaces so that should one fail, the other HBA can path the virtual host server to the SAN to maintain the connectivity.

Third party tools for efficiency
With any virtualization environment, you will need tools to do your job the best way possible. So, take some time to invest in some tools that will aid you in populating your virtual environment, specifically a physical-to-virtual (P2V) conversion tool. There are two main options available in the P2V space, PlateSpin PowerConvert and the VMware Converter tools. PlateSpin's benefit is that it will let you convert from a broader scope of physical and virtual platforms. On the other hand, if you are using a VMware virtualization solution the VMware Converter generally fulfills all needs adequately. These tools will not only save you time, but allow you to correctly size out the destination drives -- which usually makes the drives smaller -- for the new virtual machine to not waste any of the expensive SAN storage, if used.

Once you have determined the scope of your virtualization implementation, these tips will help you maximize the return on your investment while not over provisioning the host systems simply to accommodate all of the components.

About the author: Rick Vanover is an MCSA-certified system administrator for Belron US in Columbus, Ohio. Previous roles included software engineering roles with Siemens and Dematic in Grand Rapids, Michigan, working with complex supply chain execution systems. Rick has been working information technology for over 10 years and virtualization technologies for over seven years and has been publishing articles online for over six years.

Dig Deeper on Server hardware and virtualization

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.