Any technology professional with some skin in the game will tell you that once the core features of a product are realized, you then get to the extra components that don’t necessarily change the product itself. It’s similar to the candy coating of an M&M -- there are many different colors, but none of them matter to the taste of the chocolate.
This is also true now that the virtualization market is maturing, with the candy coating being the latest hypervisor features and third-party utilities promising ease of use, simplified management and the performance of a pole position race car (putting you in the driver’s seat). In the end though, what features really matter most?
As far as hypervisors go, they are all pretty similar at this point. Sure, you could argue the finer points of memory utilization, host-level security or hardware support, but if you want to implement a particular hypervisor there are plenty of vendors to choose from.
For a vast majority of those who need a virtual host, all of the solutions from the major players (Citrix, Microsoft and VMware) are appropriate and will work fine -- although the marketing rhetoric these days is over-heated and takes on an air of Mac versus PC. But while it may be a fun argument to have, the functionality remains comparable.
With this in mind, consider any specific needs for your virtualization environment and match up those feature sets. There will always be something new, but often the key differentiators are tiny. Cost is the big consideration, however. Every hypervisor is free at this point, but the cost will lie with your needs further up the stack, beyond the basic role of virtualization on a single server. The trappings of a lab environment for virtualization often don’t allow for clarity of your needs beyond a few servers and a single set of storage. Therefore, it’s important to think big because once you pick a direction it will be hard to switch later on.
It’s difficult to find server hardware that doesn’t consider virtualization to be its ultimate role. Although some vendors are pushing their blade solutions as the ultimate in integration, you should consider the density not only of servers, but of processor cores and memory. For some hosting scenarios, you’ll want as much virtual machine density as possible. This is certainly true when looking at servers for Web services, infrastructures such as DNS or Active Directory and desktop virtualization for end users. It’s also where you may want to consider larger servers that hold plenty of memory and four sockets of quad-core processors.
In terms of networks, servers are often shipped standard with four network interface cards (NICs) embedded into the motherboard. Any less and you are not following network virtualization best practices. Remember that you are supporting what amounts to software switches in your virtualized hosts. If you think you will want a primary network for your virtual servers, a management network connection for your host or parent and another when sharing storage and failover, such as VMware VMotion or Hyper-V Live Migration, you’ll want dedicated bandwidth for that purpose as well.
If you follow the best practice of having two NICs in a failover configuration, you’ll need six NICs for this purpose. Now add any other networks you may want to support for different subnets utilizing VLANs and you may need even more, or you may need NICs that support trucking protocols to send all that VLAN traffic from a set of NICs to the switches. In this case, buy good NICs and buy plenty of them.
Don’t forget that using up spare capacity may mean disaster when you need to fail over. All too often you’ll see available hardware capacity designed for host failover scenarios used up for additional virtual machines. Stick to your design documentation and stick to your guns. When something goes wrong on the server, you’ll need that capacity to keep everything running. As any systems administrator knows, you may have all the redundancy you can buy, but something will eventually go wrong.
The management stack
If the hypervisor is free, then how are these companies making money off of virtualization? The answer is the management stack -- from VMware’s vCenter line to Microsoft’s System Center Virtual Machine Manager (SCVMM) and everything else in between. Keeping track of your virtual environment quickly becomes a priority, and you’ll need the right toolsets to do it. Some vendors are selling integration with change management, release management, disaster recovery or sprawl control -- all attractive feature sets for many administrators, but they come with an added cost.
To assess your environment’s demand, first determine how much of your need is covered by vendor toolsets and then compare them to third-party alternatives.
To assess your environment’s demand, first determine how much of your need is covered by vendor toolsets and then compare them to third-party alternatives. This is especially true for any organization that is deploying more than one vendor’s virtualization solution.
Data center virtualization is one of the next great challenges. While there are many benefits to converting as many resources as possible into the fold of the virtualized data center, not all items can be integrated smoothly. Different vendors permeate many data centers, and various switches, routers, storage and servers occupy different types of functions and needs. If your equipment isn’t from an “approved partner” then this integration can get messy. Don’t spend your money on this kind of effort until you have fully vetted the solutions. Promises of full-scale virtualization can be broken without the proper planning and consideration of a vendor’s go-to partners.
New to the management stack is cloud integration, which has been the buzz for months now as CIOs start to really pick up on the trend thanks to a thorough marketing job by the manufacturers. I’ll be the last to turn away cloud integration features since cloud is quickly becoming a standard solution in some areas, but often times this integration with your virtualization platform comes at a pretty hefty price. In addition, it can limit you to specific cloud providers willing to allow for hybrid/private/public cloud integration with the tools you select and a licensing cost to them.
That may not be a big deal if you are going to use the features fully and need that integration, but remember that you are not limited to vendor tool integration for your cloud selections. Nothing is stopping you from deploying your own servers in the cloud and taking a VMware virtual machine and converting it to an Amazon EC2 type, for example. Integration can be a great feature, but unless you plan on having a good deal of cloud presence you may want to save the cost and see what you can do on your own.
You can follow SearchWindowsServer.com on Twitter @WindowsTT.
ABOUT THE AUTHOR
Eric Beehler has been working in the IT industry since the mid-90's, and has been playing with computer technology well before that. He currently provides consulting and training through his co-ownership in Consortio Services, LLC.
This was first published in March 2011