Do you really know all your space and power options?

Executive Summary

At a faster rate than ever before, companies are running out of Data Center space and power. In October of 2007, the Gartner Group predicted that more than 70% of the Data Centers in the United States will have “tangible disruptions” because of energy and floor space constraints in their Data Centers by 2011. Historically, legacy Data Centers were built with a design specification of 100 to 150 watts per square foot.  Today, these centers don’t have the power and cooling capacity to meet the needs of the current high-density computing gear. Companies are now seeing the need for 300 to 400 watts per square foot. That is expected to rise to over 600 watts per square foot by 2011 according to the Gartner Group. (1)

In 2007 IDC noted that the total power consumption of US servers was 0.6 percent of overall US electricity consumption. When cooling equipment is added, that number doubles to 1.2 percent—the same amount used by color televisions. This trend will also jump by 40% by 2010 (2). Does your family have a color television on right now?

CIOs across the country are in the throes of addressing this issue. Many are asking themselves, “Do we refurbish established sites, build new, or look to a collocation or managed services vendor?” Instead of trying to answer these questions themselves, CIOs should be soliciting unbiased third-party analysis that clearly points to the correct solution. Without an unbiased third party recommendation, CIOs could be influenced by vendors trying to sell products vs. providing an optimal solution.

This process reviews the space and power problems all CIOs are, or will, be facing in the years ahead. In addition, the process reviews options for expanding space and power; structured methodologies that generate actionable analysis results; and sets the baseline for expectations for the project.

Data Center Space, Power, and Carbon Cap & Trade Challenges

There is no shortage of articles talking about running out of space, power, cooling capacity, and the Carbon Cap & Trade taxes on the horizon. If the result of such legislation was a carbon tax cost equivalent to $50-$200 per ton of CO₂ emissions, the annual cost for 1,000 commodity servers would be $200,000 to $800,000. In his February 24, 2009, speech to Congress, President Obama addressed energy efficiency and called for Congress to send him “legislation that places a market-based cap on carbon pollution.” (3)

 

Gartner says that as the power requirements of IT gear continue to grow, energy costs will become the second largest operating cost in 70 percent of the data centers worldwide by 2009, following the cost of people, and well ahead of the cost of the IT gear itself. (4)  What is the status of your Data Center?

Green computing has its own set of challenges. Virtualization and blade server technologies will help to save on space but they will require more power per rack. A blade server is simply a rack turned on its side and miniaturized. The racks systems software is internal so that a blade or collection of blades looks like a standalone system or a system that has been virtualized with partitions. The issue with blade servers is that there is no standard form factor for a chassis and a server. So, what does this mean?  When you buy a PowerEdge from Dell TM or a BladeSystem from HP, you are buying into whole system management architecture, as well as a very specific hardware architecture.  And, it will cost more. CIOs need to weigh power costs against space costs and against the impact the IT industry is having on the environment.

Even More Criteria to Consider

Most experienced CIOs know that planning for additional Data Center capacity is not an overnight proposition. Today, it takes more time because there are many criteria to measure, and workable options can be difficult to find. Demand for backup generators, options to use diesel tanks, tougher zoning and storage regulations are but a few of the intertwined issues. The ability of a single or even multiple utility companies to provide power can be a constraint, sometimes requiring the investment in new power sub-stations.

Tier1 Research calls the wait for large diesel generators the “long pole” in a Data Center build.  They cite that large major vendors for long diesel generators have wait times of 12 to 18 months in their recent Midyear 2007 Internet Datacenter Supply Report. (5)  Data Center Knowledge.com reported large generator provider Cummins projected an “extremely strong” 2008 supported by strong demand from the Data Center sector. (6)  All of these issues are also driving an increased utilization of commercial Internet Data Centers.

In conjunction with power and space challenges, Sarbanes Oxley and the aftermath of Hurricane Katrina are driving the larger enterprises to focus on maintaining a business continuity site. As a result, growth in the global multi-tenant Data Center market has outpaced supply in the last year as noted by Tier1 Research.  Demand was up 12.53% on a global basis, and supply was up just 4.23%. Experts predict utilization trends will continue upward for 3-4 years with demand outpacing supply. (7)

Succinctly stated, there are many questions to ask, and many answers needed in order to successfully prepare for a new or updated Data Center.
 

A Structured Analysis is Key for Success

 

A methodology has been developed to help you answer these important questions.  Some of those include:

 

•  Is my analysis unbiased?

•  Do I have a strategy for replication and disaster   recovery?

•  Can my environments be geographically distributed?

•  Do I need a different tier of availability among my production, non-production, and back office systems?

•  Do I really need a Tier 4 Data Center?

•  Have I gathered ‘apples to apples’ data comparisons?

•   Have I looked at all the alternatives?

We are not saying you should discontinue efforts to pursue greener computing approaches. We are saying take an inventory of what you have today, divide the systems into the appropriate availability buckets, develop the costs to host, and run each of those groupings across the scenarios appropriate to your environment. Build any go forward programs for greener computing approaches into whatever approach you decide on for your enterprise.

Our methodology utilizes the following steps:

 

The process is simple and rapid. It starts with developing a Plan. Project management practices are employed to organize the project, ensure progress, and manage the expectations of stakeholders.  Early in the process, a workshop is held with the stakeholders to discuss the initial Pros and Cons of all the alternatives for expanded Data Center capabilities.  This workshop environment allows for obviously non-viable options to be eliminated early and defines a clear set of options to gather detailed cost data.

 

The next step is Data Collection, both the cost/metrics around current operations and projected costs for each go forward alternative. This step requires that the right level of data is gathered for each option so that a true ‘apples to apples’ or side by side comparison can be built. During this step we utilize historical data and industry standard estimating mechanisms like the revised model from the Uptime Institute, which for new construction looks at dollars per square foot plus the dollars per KW. Even items such as cost of power, tax ramifications, and potential personnel reductions that an alternative might drive are taken into consideration.

 

 

The next step is Modeling.  This step is where prior experience is most important.  The key to successful modeling is to build a model that has reference-able data and viable analysis capabilities for large and small ticket items.  We help answer questions such as:

•       Are new equipment purchases included with the right amortization schedule? 

•       Do personnel costs reflect true salary, benefits, and overhead? 

•       Are taxes treated consistently across the alternatives?

•       Are consistent disaster recovery approaches accounted for in each alternative?

•       Does a collocation alternative include multiple micro-cleanings per year 
    and the build/own/operate does not? 

•       Is an inflation COLA being allocated consistently across the scenarios?

 

A sample cost comparison chart is below.

The fourth step is Analysis. This involves the iterative process of working with the client decision makers to understand the various alternatives and ensure the data is correct. It is an iterative process because each stakeholder must understand the details of the alternatives. At this step we ask each stakeholder to grade each alternative on a number of categories beyond cost since the final decisions are not made on cost data alone. It is in this phase where we solicit a financial representative from the client to become familiar with the cost model so that later modifications can be made as needed.

Frequently during this phase there are requests to look at the data from different directions, such as, ‘How much capital is required in any one year for an alternative?’ Note below: DC refers to Data Center.

The last step is the Recommendation phase. This is not always making one recommendation; but rather helping the decision makers understand the ramifications of the choices at hand. In other words, making sure every individual stakeholder understands the data within each option, the benefits, and the challenges. Decisions are not always made based on cost alone; but rather consider corporate strategies and directions, such as green initiatives.

Standard Data Center Options

When it comes to looking at expanded Data Center capabilities there are a lot of choices. For some, the size and reliability of their Production systems is so critical to the success of their business there is a need to construct their own Tier 4 Data Center. For others, the cost of or the technical skills of their internal staff makes a Managed Services approach more attractive. Still others may have the need for a raised floor capability to look at technologies or to develop software, but their production systems are better located at a Tier 4 collocation facility. The following reviews standard Data Center options:

 

1.    Build a new Data Center: This is the first option many internal teams think about. The usual issues are the large up front capital costs and what size to construct. Size and capability questions must be answered for immediate and future needs. For example, what tier of reliability is needed? The Uptime Institute projects that a Tier 4 Data Center will cost approximately $212/ft for the computer room component and $22,000 per KW of useable UPS output. (9)

2.   Expand the existing Data Center:  Expansion always brings up two questions. Are we putting too much of our capabilities in one site? Is it possible to build a new server room and the mechanical equipment without disrupting the existing capability?  These questions need to be answered.

3.    Retro-fit an existing site:  With M&As and consolidations taking place frequently, there are often times existing facilities that can be acquired and retro-fitted to a client’s requirements. We like to consider this alternative since it frequently is less expensive than building on a green field.

4.   Multi-Tenant Collocation provider: There are several national and global collocation providers and many regional players. They offer varying tiers of reliability. Providers offer space within a rack, by rack, by customer private caged areas, or walled-off private rooms. They provide what commonly is referred to as ‘ping, power, and pipe’ services; that is, network conductivity, power infrastructure, and cooling capabilities. Often they offer additional services such as troubleshooting hardware issues, equipment repairs, or installation of new equipment. 

5.   Managed Service Provider: There are numerous Managed Service Providers. For purposes of a completely open process, we are using the ‘up through the operating system’ as the definition of a Managed Services Provider.  For this option, the managed service vendor provides all hardware management, as well as management and support for the operating system. The application layer is still managed by the client. 

6.    Full Outsourcing Provider: Many of the larger outsourcing organizations provide the Data Center services to take over not only hosting your hardware within their Data Centers but the management of the operating system layer and the application layer as well. This is common for ERP applications from providers like SAP and Oracle. Application management does not stop at just packaged applications. These vendors will learn how to operate and maintain your custom built applications, too. This is the diametric opposite from building a completely new Data Center. It is transferring the responsibility for your applications. This type of outsourcing alternative usually requires gathering detailed information and soliciting custom, in-depth
proposals from several vendors.

Many times, organizations opt to select a combination of alternatives from each scenario. Given the number of criteria needed for a thorough analysis, it is simplest to divide a Data Center project into two categories:

 

1.      Production and back office systems that require higher availability and a service level for that availability.  These are the customer facing and revenue generating systems where the availability has a real financial impact on your business.

 Non-production and non-critical back office systems which don’t require as high an availability level and usually do not have the same disaster recovery requirements.  Yes, a business is impacted when these systems are not available, but there is often a viable work around or other activities that can be performed when these systems are down. 

 

“Divide your systems into two tiers to allow the right
availability to be achieved at the right price.” 

Putting It All Together

Once an analysis has been completed, a detailed financial model with side by side comparisons is revealed. An unbiased analysis gives decision makers both numerical and non-tangible data to evaluate the best path for the company moving forward. 

Highlights

·   When pulling together alternatives, it is best done by an unbiased third party who does not benefit if one solution is chosen over another.

·   Keep in mind, even internal teams may have agendas for a ‘go forward strategy’ that could impact receiving an unbiased analysis.

·   Flexibility is key. If breaking your systems into tiers for a more effective analysis, then splitting alternatives should be done as well.

·   It is important to determine both tangible and non-tangible benefits and risks.

·   There are a lot of changes coming with the Carbon Cap and Trade legislation, and your analysis needs to be aware of the direction it is taking.

·   To avoid a time-consuming analysis provide a deadline for the analysis to be completed.

·   There is great industry projection data available. Find a resource that knows where it is and how to use it.

 References

(1)   Timothy Morgan, 11 October 2007, The Unix Guardian

(2)   Jonathan Koomey , Feb 2007,  Professor at Stanford University with  reference to IDC Study Data

(3)   See President Obama’s February 24, 2009, remarks to the joint session of Congress.

(4)   Timothy Morgan, Dec 2006, The Gartner Group

(5)   Daniel Golding, 2007, Internet Datacenter Supply Report, Tier1Reseach

(6)   Data Center Knowledge.com

(7)   Daniel Golding, 2007, Internet Datacenter Supply Report, Tier1Reseach

(8)   Data Center Knowledge.com

(9)   W. Pitt Turner, 2006, Dollar per kW plus Dollars per Square Foot are better Data Center Cost Model than Dollars per Square Foot Alone