Many vendors and IT consultants today tout server consolidation and virtualization as the key to cost reduction and operational efficiency, and the best way to meet the evolving support requirements of data center facilities, network and telecom infrastructures, and disaster recovery plans.
But too much focus is being given to server consolidation – and to server virtualization specifically – as an end in itself. A more reasoned and effective approach is to view server consolidation as one component of an effective resource optimization strategy.
Here’s a few of the most common misconceptions surrounding server consolidation:
Misconception #1: Server consolidation and virtualization save money by reducing personnel requirements. Server virtualization does not reduce the number of logical servers. In other words, while three physical servers might be consolidated into one physical server hosting three logical partitions, people are still needed to manage every virtual machine instance, associated application, and database. This often increases complexity, so the new environment can require the same or greater levels of management than before.
Process changes are also required in consolidated environments. Since more of the company’s critical applications will be running on fewer servers, capacity planning will be more important than it ever was before. Data centers will need to look to their mainframe operations to find procedures that should now be emulated in the world of consolidated midrange servers, and this new level of on-going, proactive planning takes time. Certainly consolidation changes the nature of management tasks, but it doesn’t necessarily reduce the overall effort
Misconception #2: Server consolidation saves money on software licensing
Not necessarily. While basic virtualization software is free, most companies run more advanced software that requires additional licensing costs. Software vendors are also reexamining and altering licensing models that are currently priced on a per processor or per server basis in order to protect revenue streams in the new virtualized data center.
For example, let’s say you have three separate servers with a single processor, each running a single application licensed to run on this configuration. If you move these three applications onto a single, more powerful server with multiple processors through virtualization, the software licensing model may change from pricing for a single processor solution to a multiple processor solution, and you incur the additional cost of the virtual software licensing – a cost the business will be reluctant to accept. Even if the license fees for the application itself don’t change, the fees for the middleware that the applications require may. This all has to be taken into account when deciding which applications to run on which servers. It’s worth noting that in some cases software vendors are adopting some flexibility with their licensing models to accommodate these situations.
Software and personnel savings can be increased by standardizing the environment and reducing complexity as part of the consolidation initiative (i.e., reducing the number of operating system versions, eliminating redundant software tools, etc.). Many claims of savings on software costs for consolidation projects are actually a result of the standardization that took place at the same time. A critical success factor is support from the senior leadership team in driving to a standardized environment.
Misconception #3: Virtualization means increased reliability. Server virtualization puts “more eggs in one basket” so that one outage will affect more processing and more users. As the number of production servers is reduced, the requirements for redundancy increase in tandem. Critical applications can share servers in a high-availability configuration, but the largest, most critical applications may still warrant dedicated hardware.
That said, application consolidation – where a single application that requires multiple servers is consolidated onto a single physical server without virtualization – is always a good idea as long as that single server can provide adequate performance. In addition to saving hardware costs, this also improves reliability by reducing the number of points of failure for that application.
While it may be possible to leverage development teams and design applications to run on fewer servers without the need for a partitioning product (virtualization), developers are not always cooperative, and even when they are, projects can move quite slowly. This means that benefits from application consolidation can take a long time to be realized. In one recent example, a large state government initiated the consolidation of its 300+ email servers into a single server. While the initiative eliminated many points of failure and made the system easier to manage, the state did not plan or account for a disaster recovery server, nor anticipate the need to upgrade the entire network system to accommodate increased bandwidth requirements.
Misconception #4: Hardware and software vendors will be objective. Organizations need to exercise caution when choosing their partners in a consolidation endeavor. Hardware vendors, in particular, see server consolidation as a significant market opportunity. The current trend to encourage companies to purchase new, high-powered servers that can manage an increased number of applications ignores the fact that utilizing existing idle capacity produces the highest savings.
Put differently, a 64-processor box with 16 partitions costs considerably more than 16 quad processor boxes, and the large box yields higher margins for vendors).
Vendors may also try to protect their revenue stream from the client by influencing the consolidation solution itself. For example, in one case vendor A indicated the client could consolidate on a 2/1 ratio using their hardware; vendor B indicated the client could consolidate on an 8/1 ratios using their hardware; and vendor c (an independent consolidator predicted a 25/1 ratios with either vendor A or B’s equipment.
Misconception #5: Server consolidation and virtualization is a one-time IT project. Many organizations undertake consolidation initiatives thinking it’s a one-off project; this lack of consideration of long-term implications can lead to serious problems. Because virtual machines are so easy to implement, preventing server sprawl and “virtual sprawl” of the virtual machine should be top of mind when defining the initial consolidation strategy. The planning that goes into the initial project has to be documented, proceduralized and developed into a core competency.
The power and cooling requirements of the data center should also be considered. While more processing and storage capacity can be packed into data centers these days, the facility is now often the limiting factor. On average, it takes two to three years to design and build out a new data center facility and orchestrate the move to the new facility. Ongoing management processes for the data center will certainly change with a virtualized and consolidated infrastructure; so too should decision processes surrounding the infrastructure to avoid the occurrence of any of these problems.
Misconception #6: Virtualization is a strategy. Ideally, the starting point should be defining an overall strategy for the data center. Too often companies dive down a particular path without considering the bigger picture of the operation. A “resource optimization” strategy that incorporates a company’s entire IT operation can provide an organization with visibility into the most appropriate strategies given its unique environment.
The idle CPU capacity that exists in the data center may be best used to provide failover capacity for a disaster recovery scenario, or to provide distributed processing for critical applications at specific times of day. The decision on how to optimize resources will vary dramatically from company to company. From this perspective, virtualization is just one part of an overall approach to resource optimization that includes many considerations (including hardware sizing, software licensing, storage architecture, capacity planning, facility issues and network management) and options (including active/active disaster recovery solutions, distributed processing and, in many cases, consolidation without virtualization).
Business Support: the Key to an Optimized Infrastructure
The real key to an optimized data center operation is application management. Since applications are for the most part controlled by business units and supported by the IT infrastructure team, the latter’s influence is by definition limited, as applications drive the infrastructure requirements.
Put differently, solutions such as Storage Area Networks and server consolidation cannot be fully leveraged in a non-standardized or antiquated application environment. Unless a business unit or senior management is willing to move the organization to standardized applications that run on a standard set of hardware and operating system(s), the infrastructure team will be stymied in optimizing the infrastructure via consolidation, and the full benefits will not be realized.
Ultimately, applications define the parameters of the infrastructure. Organizations that focus on hardware as the “solution,” and decline to tackle the central issue of application management, are seriously limited in their approach to resource optimization.
Scott Feuless is a Compass senior consultant based in Houston, Texas. Compass (www.compassamerica.com) is a management consulting firm specializing in operational improvement for Fortune 1000 organizations.
