By Vijay Swami
There are many aspects to virtualization in an Enterprise. Virtualization can be viewed as a way to reduce the siloed segments of infrastructure in your company and, instead, turn your entire infrastructure into a shared resource. In this way all applications, departments, projects and people can have virtual access to the entire infrastructure. Considered in this light, a virtualized Enterprise may seem like nirvana. There truly are significant CAPEX and OPEX savings possible for Enterprises that effectively deploy a holistic approach to a virtualized infrastructure. This article will help you evaluate the best way to accomplish this in your own Enterprise. By evaluating each potential area to be virtualized, you can determine the best way to deploy this approach to achieve the maximum benefit.
The Server Perspective
By implementing server virtualization, an Enterprise can benefit and effect savings in a number of ways. For example, server virtualization reduces the server footprint in the data center because multiple physical servers are consolidated onto fewer hosts. This consolidation reduces the required amount of space, power and cooling and, in the long run, reduces the need for capital spending on future servers. Enterprises that deploy server virtualization can often see a consolidation ratio of 20 to 1, which can amount to huge cost savings overall.
A second benefit of server virtualization relates to reduction in server downtime. In traditional data centers, each application runs on its own physical server. If that physical server fails, there is no built-in method to assure high availability and it can take hours or days to recover after a failure. In a virtualized server environment, if a physical server fails, the leading hypervisors offer the ability to automatically restart virtual machines on other servers in the cluster. This means that data centers using server virtualization can recover in minutes. Since downtime for an application can cost Enterprises several hundreds of dollars per minute, the significant hard dollar savings by minimizing unplanned outages may be the single justification that an Enterprise needs in order to implement virtualization.
The Desktop Perspective
Enterprises have two issues related to their desktop environments – management and maintenance. Each physical desktop is subject to frequent failures of parts that may have become infected with viruses or have misconfigured OS images due to user interaction. Enterprises pay a significant amount to fix these desktop failures. They need support staff to fix the problems and they lose the desktop user’s productivity while their system is being repaired. It became very clear as the virtualization industry matured that the desktop space would be ripe for this type of evolution.
Enterprises that deploy virtualized desktops actually run the user desktops as a virtual machine in the data center. The user end point is simply a thin client. With this configuration, Enterprise IT can patch and maintain consistent OS images from a centralized location. They can deploy applications centrally and easily. They can carefully manage access to specific applications by various groups, ensuring security and consistency throughout the Enterprise.
If one of the thin clients fails – which, in itself, is rare – it can be replaced in minutes so the user does not lose valuable work time.
Therefore, if you view management and maintenance as the two biggest issues of desktop environments, then a virtualized desktop essentially solves both handily and provides excellent payback for the investment. Additionally, Enterprise IT can focus its energy on innovative desktop technology that will benefit the company rather than spending unnecessary time repairing and updating physical desktops.
The Storage Perspective
In a more traditional Enterprise environment, SAN storage is managed by taking a set of disks, assigning them a RAID protection level, and carving storage space out of that RAID group. With this approach, however, there can be wasted storage capacity that can lead to overspending on storage infrastructure. For example, if a project requests a terabyte of storage space – which could be twice as much as needed, the space will be fully utilized on the array and never be able to be reclaimed, even if the project never needs it. If this process is repeated with multiple projects, the problem of wasted space is exacerbated and storage reclamation is a non-trivial task for Enterprise IT to tackle. When other users request storage, IT will need to purchase more even if there is unused storage space available.
In a virtualized storage configuration, the entire storage array is presented for usage and the administrator can abstract disks and RAID groups as needed. In the project described above, the administrator could assign 1 terabyte of space to the project, but the disk space on the array will only be consumed as it actually is used by the application. The existing storage capacity of the Enterprise is set up to be used much more efficiently with this approach. This will help the Enterprise reduce its storage acquisition initially, and will also help the Enterprise to avoid buying more storage than is needed because the available storage is difficult to reclaim.
The Network Perspective
Just as disks in traditional arrays are tied to particular applications, this physical coupling approach also plays out in the network. This means that fiber channel switch infrastructure, IP data switching, management switches and backup traffic switches are all separate from each other. This type of siloed infrastructure creates problems because switch ports are not consumed symmetrically. For example, if there is a need for more storage switch ports, but only IP data switch ports are available, it does not work.
In a virtualized or converged network, the Enterprise can deploy a new Data Center Ethernet as well as a new Fiber Channel over Ethernet protocol. Together, they enable one 10 gigabit switch port to pass both fiber channel storage and traditional IP traffic, while allowing the segmentation of the traffic to occur via software. In this way, the Enterprise IT staff does not need to purchase one switch for management, one for production data and one for fiber channel. Instead, they can purchase a single switch capable of doing all three functions and more via a single 10 gigabit Ethernet port. Software can be used to partition the traffic types and bandwidth.
Shifting to a virtualized network completely eliminates a siloed switch infrastructure, offering huge savings in capital and operational expenses with the reduction in the number of required cables, switches, switch ports and management resources in general.
As you evaluate the costs and benefits of a virtualized approach, it is helpful to review a checklist that allows you to consider virtualization as a holistic approach to data center design, and also examines its impact on each critical component of your IT infrastructure. With this type of careful analysis and consideration, you will be best equipped to choose the right strategy and products to achieve a virtualized infrastructure that immediately realizes the benefits and cost savings that make this new approach so appealing.
About the Author: Vijay Swami is the Data Center Technology Manager at Presidio Networked Solutions (www.presidio.com).