Twenty-four North American data center executives from finance, hosting and third-party data center providers, retail, health care, manufacturing, transportation, designers, and facility operations helped draw a picture of the future in recent interviews. These individuals are all responsible for data center site infrastructure and some have additional information technolgoy (IT) duties. Individuals in the Uptime Institute and The 451 Group offered additional perspectives.
Figure 1. The Uptime Institute breaks down data center costs, with IT capital costs being the largest segment.
According to the executives, tracking and managing total cost of ownership for data centers will be one of the top challenges facing data center owners and operators in the coming years. Managing data center and IT costs holistically has been a challenge, as costs are widely spread across multiple organizational boundaries, making it a difficult and complex task. And without correct, complete information, data center owners and operators can make misguided decisions.
Jonathan Koomey’s 2008 technical research paper, “A Simple Model of True Cost of Ownership,” provides the most complete methodology to incorporate capital, operating, energy, and other costs necessary to support IT. Uptime Institute expects to see the TCO approach (see figure 1) more widely used and the only acceptable methodology for the majority of the industry. This may be accomplished through the use of ‘management accounting’ (collecting information relevant to a decision from any appropriate source) when the official ‘financial accounting’ systems are not responsive.
The availability of capital is another financial consideration for data center managers. In 2006, enterprise firms had access to capital, given a sufficient justification story. Third-party firms were still dealing with the surplus of data centers on the market and had difficulty raising capital. Today, it has reversed, and Uptime predicts that in 2016, many enterprise teams will find capital is no longer available for data center investments.
In 2006, the primary business focus was on availability and reliability. Today, operating costs have forced many companies to focus on reducing energy consumption and capital avoidance. This is an appropriate direction for many data center operators. Mike Lewis, director, mission critical engineering at eBay commented that availability and reliability were the principle focuses for eBay’s first data center. eBay focused on energy efficiency and overall sustainability at its newer site while maintaining high levels of availability.
Figure 2. Simple definitions of the Tier System.
In the future, more IT availability requirements will be met through the design topology of the IT systems. This does not mean reducing the use of Uptime Institute Tier Standards for building topology, since that is a common language that has been adopted internationally and is widely used. In a recent presentation, Scott Noteboom, vice president, operations for Yahoo! used Tier references extensively as he outlined his vision for reduced cost and energy for future Yahoo! sites (see figure 2).
Any discussion of future performance would be incomplete without a mention of scalability or modularity. By 2016, wider use of manufactured solutions will be evident. Regardless of how power, space, and cooling requirements will be met, it is clear that most sites will meet the business needs using more building blocks of smaller capacity.
The increasingly common use of more and smaller blocks for enterprise uses raises the issue of risk tolerance. Multiple blocks of capacity supporting an integrated set of IT solutions may, in some instances, actually increase the risk of disruption since a failure in any of the small blocks may impact the logical interconnection that exists within an IT architecture. This risk needs to be identified and more clearly addressed in order to be resolved. This is a requirement for success in every site as greater scalability is adopted.
Over the past five years, the data center industry has developed many useful new best practices. While many data center operators have adopted some of these,
Figure 3. Uptime Institute survey results show that the largest group of organizations measure PUE at the UPS outlet.
very few have taken a holistic approach and adopted all of the latest best practices. According to the Uptime Institute’s 2011 Data Center Industry survey, only 12 percent of respondents are managing supply air temperature at the server inlet, the most accurate location for managing server cooling. Data center managers controlling temperature at the cooling unit return or room level are most likely over-cooling their server hardware and wasting money and energy.
In the coming five years, data center operators will need to be more aggressive about adopting these best practices in order to compete economically.
Management or organizational cultures are the obstructions to implementation of good practices throughout the industry. There are well-developed technical solutions (airflow management, increased computer room temperature ranges, IT consolidation, etc.), but they have not been adopted widely in all sites. The barriers are management, risk tolerance, perceived risk, and so on, rather than a lack of technical solutions.
The increasing public pressure to reduce energy use comes with emerging regulatory ‘assistance’ for the industry. This unwelcome assistance could be avoided if the industry took the initiative and widely implemented the established opportunities to drive up IT productivity based on work per unit of energy. The focus the past five years on improving the power utilization effectiveness (PUE) has generated notable savings (see figure 4). But the industry needs to realize that once a PUE falls below 2.0, the preponderant use of energy is the IT kit. With a PUE of 1.5, 67 percent of the site energy use is from IT uses while only 33 percent are from the site infrastructure supporting IT. Metrics that compliment the PUE and focus on the IT systems are sorely needed.
Many actions unthinkable a few years ago are now becoming mainstream, even with enterprise sites. Locating the site where low/no carbon power is available,
Figure 4. And the most common PUE is in the 1.6 to 1.79 range.
where the cost of power is low, and where the outside air can be used for free cooling were not available options when the data center needed to share the postal code with the CIO’s office. Internet companies started that trend, and it will be widely adopted by 2016.
An operational consideration often ignored in the industry is the aging work force in facilities operation. The industry needs to find a source pool of well-trained and talented individuals to replace the existing work force. Part of the solution may be development of comprehensive training courses focused on the data center industry. Outsourcing is not a solution; it simply moves the location of the deficiency to another company, which still needs to solve it.
Five years ago, a substantial rift often existed between enterprise IT and the corporate real estate organization responsible for designing and building data centers. Real estate drove the project because it was considered a building. That thinking has changed in many companies, with the realization that the building is really just another piece of the IT kit. Leaders in the industry now have IT leading data center projects, using corporate real estate teams for their skill sets of project management and construction and reserving the key decisions for IT. Successful companies will continue to follow this trend. Those who do not understand that buildings are part of the IT kit will become significantly disadvantaged.
The integration of procurement with IT is another area that has shifted and will continue to do so. IT organizations that carefully built a dual-power path site, at a substantial cost, were frustrated by procurement decisions to save a few dollars and buy single-corded IT devices. That pattern represents classical sub-optimization driven by poor incentive structures in divergent parts of the corporation.
Figure 5. Almost all supply air temperatures are in the range between 65 and 70F.
Bob Cashner, senior vice president, corporate properties group of Wells Fargo Bank, refers to the three-legged stool thusly, “IT, real estate, and procurement. It takes all three working together to be successful. Without that cooperation, they all fail. That is not acceptable to us.” Wells Fargo (including many team members from Wachovia) has figured out how to integrate its teams. Those who have not, or will not do so, will fail. It is really that simple.
The notion of a three-legged stool is simple but profound. One of the dangerous trends Uptime has noticed is the increasing number of sites where the integration of the most important teams has gone backwards. This is distressing and needs to be corrected.
Years ago, it seemed the use of hosting or third-party services was binary: an enterprise firm did or did not use them. Today brings an increasing trend to see hybrid IT solutions. Virtually all IT solutions will be a mix of owned data center, cloud services, and hosting or third-party services. Of course, managing this mix will present new learning challenges for the management team. For some data center users, the owned data centers may simply disappear, being replaced by cloud, hosting, and third-party services.
This all amounts to a complete re-engineering of the way data center services and space are delivered. Part of this re-engineering needs to be new and integrated solutions developed jointly by facilities and IT manufacturers. These trends may present major issues to traditional suppliers of data center design and construction services. The pre-engineered modular solutions, repeated application of smaller capacity blocks, and fewer (but larger) cloud service providers pushes traditional design engineering and construction toward a commodity.
Taken as a whole, the Uptime Institute research lead to the following summary of the next five years in the data center industry in North America.
• Total cost of operation will be the singular driver to data center decisionmaking. TCO needs to include all costs and be used to drive all decisions. If a piece of data is key, it needs to be included.
• Data center delivery will be re-engineered. Hybrid solutions of owned, cloud, host, and third-party will become commonplace. Traditional engineering and construction will be pushed toward commodity services. IT service reliability will shift to IT-based topology.
• Data centers and the Uptime Institute Tiers Standard are not going away. The size, location, and ownership of data centers may be different, but data centers are part of the future. Forever.
• Metrics to compliment PUE and focus on the IT side of the consumption model are absolutely necessary. Otherwise, the industry is failing to focus or measure on the largest user of data center energy: the IT kit.
• Education of management and the entire team is a continuous process. Failure to educate management will lead to loss of understanding (and funding) of what the team’s contribution is. Replacing the aging work force is essential to continued operational success.
Reprints of this article are available by contactingJill DeVries at firstname.lastname@example.org or at 248-244-1726.