To virtualize or not to virtualize? Before deciding, take a close look at this technology's hidden costs.
In the rush to consolidate servers and save precious data center space, many companies are turning to virtualization. By consolidating a data center's physical servers into virtualized servers, the enterprise can experience higher server utilization and increased operational efficiency.
Although virtualization brings lots of benefits and features, it is important to keep in mind that virtualization is not the Holy Grail and may not be appropriate for every situation or environment. It does come with some drawbacks. These drawbacks may not be deal breakers, but you should be aware of them before you adopt the technology. In this column, we will look at several key issues that can cause problems in a virtualization adoption.
Part one will discuss the power and heat costs and the management concerns of virtualization. Part two will discuss networking issues and the problem of virtual machine sprawl. Finally, part three will discuss licensing, performance and storage.
With that out of the way, let's get started.
Power and heat costs
One of the first problems often overlooked is the high power consumption and high heat output of a physical server hosting multiple virtualized servers. Remember that virtualized servers cause the hardware to run at a higher utilization rate, which means the power required to operate the physical server hardware goes up, as does the heat output.
A physical server running many virtual machines may run at a constant 80% or higher utilization rate, at which point the hardware is pulling quite a bit more power than a non-virtualized server running on average around 15% utilization.
Many data centers are not equipped to deal with the new power requirements that are necessary to support virtual servers. For example, you might need four 20-amp power circuits for each 42U data cabinet that is fully loaded with standard 2U servers, in order to satisfy power needs and to provide power redundancy (the ability to withstand a single circuit failure). This power scheme is actually double the standard power layout, which provisions only two 20-amp power circuits per cabinet.
And doubling the power per cabinet poses new problems of its own, especially if it means the data center needs to upgrade its whole power infrastructure.
Virtualization also raises new management concerns, especially in the areas of patching, backup, host provisioning, security, monitoring and hardware requirement.
In a large virtualized production environment, it can be a challenge to apply software patches to both the physical host servers and the virtual machines. If system reboots are required, it impacts business a whole lot more to take down a physical machine hosting hundreds of virtual servers than it does to reboot a machine that serves as only one server.
So while in a non-virtualized environment, a company may not have to significantly plan for downtime around a Windows Update patch, in a virtualized environment, patching and rebooting a single host server creates a much more significant impact. Planning downtime around something as simple as patching becomes an arduous task. Companies running large enterprise environments with hundreds of virtualized host servers face a complicated maintenance window, leading some to apply only the most critical patches which leaves them in a more vulnerable state.
Some companies, such as Internet Security Systems (ISS), offer virtual security patch technologies to protect servers without having to go through the actual patching process. This does not, however, resolve the problem of applying stability or bug patches, which have to be applied on the physical server. In these cases, reboots are unavoidable.
The bottom line is that physical server reboots can be very costly in virtualized platforms.
Implementing a sound backup solution in a virtualized environment is another major pain point. The first obvious question is what data do you want to backup, which is then quickly followed by how often? Is the answer backing up your entire virtual disk repository and all disk images? Or perhaps you should consider only backing up the base images and then the various REDO or undo disks?
Either way, you could be facing an uphill battle. The sheer amount of data needing to be backed up can be staggering. Unless the company is lucky enough to have a very robust SAN or some kind of a backup-to-disk solution, backing up many terabytes of data to tape is way too slow and inefficient, and we still haven't even considered how to go about backing up the physical host server or its operating system configuration.
Automated physical host provisioning is highly desirable and can cut down drastically on the physical server deployment time. But physical server provisioning is only half the battle. We have to consider automated provisioning of the virtual servers once the physical host server is up and running and fully configured.
Many technology companies provide an automated solution for deploying and provisioning physical servers. Many of those same companies, such as Altiris, are now capable of automatically provisioning virtual servers as well. But the process can be complex, especially if your organization has a heterogeneous virtualization environment with multiple virtualization platforms such as VMware and Microsoft Virtual Server. The additional complexity of automating both a physical and virtual environment has to be carefully weighed and considered.
Security issues are more complex in a virtualized environment because you now have to keep track of security on two tiers: the physical host security and the virtual machine security. If the physical host server's security becomes compromised, all of the virtual machines residing on that particular host server are impacted. And a compromised virtual machine might also wreak havoc on the physical host server, which may then have an ill effect on all of the other virtual machines running on that same host.
Monitoring is another critical area in which virtualization has its own set of challenges and unique problems. You will have to monitor both the physical servers and the virtual machines to make sure that your environment is fully operational, since losing a physical host server translates to losing several virtual machines.
Running monitoring software or agents on the physical host server can potentially cause a negative impact to the performance of the virtual machines by taking away valuable memory and CPU cycles that would be consumed by the virtual machines. It's important to calculate the percentage of physical resources used by the monitoring software to see if you can spare those resources.
Finally, it is not uncommon to require multi-port network adapters and several physical connections to each virtualization host server. Expect three to four network ports to be cabled on each server, especially if you are running a virtualization platform that does not offer the ability to trunk multiple VLANs across a couple of physical channeled links. This introduces yet another management problem; your environment will be consuming a lot of Ethernet cables and your switch-port burn rate will be pretty high. Expect to purchase a lot of access layer Gigabit switches to handle your virtualization hosts.
About the authors: David Marshall is a senior member of the reference architect team at Surgient, Inc., and he specializes in server virtualization, virtualization applications and Windows administration. He also runs the InfoWorld Virtualization Report, as well as the virtualization news blog, VMBlog.com. David is also a co-author of Advanced Server Virtualization: VMware and Microsoft Platforms in the Virtual Data Center, a book that details years of hands on experience using and implementing server virtualization solutions.
Dan Knezevic is a senior network engineer and a team lead for the data center operations team at Surgient Inc, providing expertise in the data center network and server infrastructure as well as virtualization platforms. He also specializes in network security and enterprise storage solutions. He brings six years of virtualization integration experience in the data center environment.