MYTH #1: Data center means the same thing to everyone. We’re all speaking the same language.
Because there is no standard “Webster’s” definition of data center, the term is open to a broad degree of interpretation. A lot of consultants and vendors claim data center expertise in broad terms, but they may actually only have true expertise in a certain aspect of data center projects. And their area of expertise may not match up with the needs of a specific project. Like an actor who is miscast in a role, vendors and consultants with the wrong kind of data center expertise can damage the success of a project.
The ambiguity in definition causes real problems when it comes to building and operating data center facilities, which requires proficiency in real estate, operations, and engineering. Partners must collectively have expertise in all three areas or have the organizational ability to get these three sets of specialists to effectively work together. Therefore it is critical to look for more than just “data center experience” – which means so many different things to different people.
What are the right questions to ask?
What is your history in constructing and operating data centers?
While the number of data centers the vendor has built and operated (the more the better) is important, the dates of their projects tells how familiar they are with the latest innovations in areas such as design, power delivery, and cooling architectures.
Can they show you more than plans?
A potential provider should be able to produce extensive project documentation, including copies of the appropriate local permits, power and contractor agreements, and equipment purchase orders.
Does the equipment actually work?
Being able to see that the required equipment is on site is certainly better than viewing a purchase order, but it is also not a guarantee of availability. A transformer in the yard is no assurance that it’s actually receiving power from the utility. The data center owner/operator should insist on verifying the operational fitness of all in-place equipment.
Where is the fiber? Unlike hand grenades and horseshoes, close doesn’t totally count when assessing a potential data center’s connectivity options. Fiber providers may surround a facility, but how many of them are actually inside? Having multiple providers within the building not only provides connectivity choices, but also eliminates the time and cost associated with working with a vendor to have them pull their fiber from its termination point into the “point of presence room”.
Is the data center commissioned? Asking this question is especially important when considering facilities billed as “move-in” ready. The commissioning process encompasses the full burn-in and detailed testing of all of the major components and systems of the data center. Completing this process is the only way to be sure that any problems that may have arisen have been corrected and the data center is ready for operation.
Many data center professionals use square footage as the standard unit to describe a site at its most basic level. Unfortunately, in the realm of the data center industry, the definition of square feet is not universal across all involved parties. For example, a real estate professional may use square footage to describe the entire rentable area (raised floor, common areas, etc.) within a building. An IT counterpart may use this same terminology as reflecting the cabinet footprint, and an engineer’s conception of the term may be the data center’s gross raised floor area wall to wall. This diversity of interpretation can lead to a variety of issues including overestimating the required power capacity leading to the development of a facility that is unable to accommodate future growth because it is literally out of floor space.
Describing IT load in kilowatts is the best and most accurate method of determining space and power requirements. Adopting this method of determining and expressing data center leads to power requirements based on the footprint of all the components that will reside above the raised floor. Data centers rarely have a homogeneous load profile. While racks of blade servers may indeed possess substantial power requirements, other components that will reside in the data center with them do not. Cabling and patch panels, for example, require no power at all, and other components such as network/telco (10-50 watt per square foot) and spinning disk storage (100 watts per square foot) have rates of power consumption considerably below those of high density servers. Thus, original power projections typically turn out to be considerably lower.
MYTH #3: “You Definitely Need a Tier IV Data center (Component Level Redundancy)”
The concept of tiers helps provide some relative scale of comparability between data centers. However, this lowest common denominator approach to categorizing computing facilities is also its greatest weakness as it applies to own data center planning. By definition Tier IV data centers must have full system plus system redundancy. As comforting as this redundancy may sound in theory, it also raises questions. For instance, does a given data center really need two fire suppression systems an additional four generators to back-up the original four-generator architecture? These redundancies have costs.
Fundamentally, the amount of redundancy required within a data center is a function of an organization’s level of risk tolerance in relation to cost and performance. In establishing the degree of risk for an organization, it is important to assess risk at the component level.
Considering factors like planned usage, mean time to failure, etc. may lead to the 2N level for UPS systems but only N+20% for HVAC units. In short, redundancy must be addressed on a topology-by-topology, component-bycomponent basis.
MYTH #4: “Don’t Worry. Your SLA Makes Sure Your Provider is Motivated to Fix Your Data center Problems”
An SLA (service level agreement) does not motivate a provider to fix a data center problem. How any problem is addressed within the facility is a function of the provider’s operational procedures and not the text of any individual SLA.
A prospective provider’s SLA should include:
• The ability to review maintenance documentation. These should be written guidelines covering everything from how often UPS batteries are checked to the set point temperature for the water-cooled chiller. The review should include an assessment of comprehensiveness of the documentation and its inclusion of parameters (set points, etc.) that are based on the defined standards of organizations like ASHRAE (www.ashrae.org) and NETA (www.netaworld.org).
• The experience of the water chillers. The evaluation of the SLA should consider the data center’s operational personnel. The document should specify the qualifications of each technician or facilities engineer who will be working on specific components. Using “generalist” or cabling technicians to service the major components of your data center may be fine for your provider-it shouldn’t be.
• Real-time performance monitoring. The provider should clearly define the tools required or available to ensure that all prescribed maintenance activities are performed on time and on schedule.
• Records Auditing. Components rarely “just fail”. Very often these failures are preceded by patterns of performance that indicate their need for component adjustment or replacement. A comprehensive plan should specify that performance data is auditable by the owner and collected on an on-going basis, and specify how it is used to aid in the prevention of component outages.
With the substantial growth of high-density applications the need for power within the data center borders on the insatiable. Many data center providers oversubscribe power availability, which means allocating more power to users than the total amount of power available. When contracting for power, always verify that your potential data center provider does not oversubscribe at any level (utility, UPS, PDU, circuit) of the components that make up their overall power architecture. Moreover, you should narrow your search to vendors who will clearly state, and commit to, a fixed kilowattage for your IT load independent of the quantity of power circuits.
And to pay only for energy actually used, the operator can insist on metered power records from the provider. Insisting on a hard capacity commitment in conjunction with metered power can avoid the unpleasant pitfalls associated with power oversubscription
Building or obtaining a data center that meets an organization’s current and future needs requires careful planning based on accurate information. It also requires expertise in multiple skill sets including real estate, engineering, and technical operations. Since not all organizations are proficient in each of these areas, focus on identifying and working with a partner that is. Making informed, educated decisions is an essential requirement to make sure that the data center performs as required. Understanding the fallacies of the myths surrounding the industry is a good first step in the planning and decision-making process.