There are plenty of theories on best practices when it comes to data centers. Below are several do’s and don’ts that the Faith Technologies team encourages all of the builders it works with to keep in mind on data center projects.
Consider alternatives to an EPO
While emergency power off (EPO) systems are required by some insurance companies, this feature is mainly a tool for systems with plenum-raised floors. If this doesn’t apply to your project, use other options.
The need for EPOs is a common misconception because of Art. 645 in the National Electrical Code (NEC), but Sec. 90.3 amends that it is not required anymore. This Code amendment is important because using EPO alternatives increases reliability and uptime of systems while decreasing issues with budgets, schedules, and headaches encountered during the commissioning process.
Good examples of providing the same level of protection as an EPO (but limiting human error) include: thermal runaway protection, selective coordination, fire protection and very early warning smoke detection and alarm (VESDA) pre-action systems tied to building automation systems with automatic notification and escalation alarms through text, email, and phone.
Build in stability and flexibility for IT equipment consolidation or movement
Overall IT infrastructure spending is up 2.8% from the previous year due to the sheer growth in the amount of data being gathered and studied by users. International Data Corporation (IDC), Framingham, Mass., forecasts that IT data center cloud spending will increase by 21.2% between 2015 and 2019, while traditional IT infrastructure spending will decrease by 4.3% (Figure).
Despite the decrease in infrastructure spending, it is still projected that $63.6 billion will be spent on traditional data centers in 2019. This means data centers will need to be flexible to accommodate new and emerging technology.
Data center owners may face tough choices regarding scaling, particularly if the facility equipment is nearing the end of its use. This could lead to a large volume of “moving” and “consolidation” type projects. Scalability happens in both directions, so consider the following in your design proposals:
• Size of the incoming utility service
• Cost of electricity
• Modularity of the uninterruptible power supply (UPS)
• Incremental size of the generator
• Utilization of bus duct instead of conduit and wire to the racks
• Generator and heat exchanger space
• Sustainability and accommodation of free cooling
Performing a design review with these items in mind can improve the success of your project.
Consider prefabrication
Prefabrication is defined as the practice of assembling components in a factory environment and transporting complete assemblies to the construction site. It is often the safest, fastest, and most cost-effective option for a project because the components are built in a controlled environment.
Building components in a controlled environment mitigates many risk factors, including temperature, position of assembly, having the right tools, and having an expert manufacturing process in place. The parts are readily at hand, building is optimized, delays due to weather are eliminated, and coordination with other trades is reduced, particularly in congested areas. All of this makes prefab more cost-effective because it takes less time and personnel to complete.
Don’t assume there is a dual-path power redundancy
When upgrading an aging data center, end-users often struggle to find confidence in their dual-path redundancy scheme, crossing their fingers when shutting down power distribution units (PDUs) while transitioning over to a newly installed system.
A successful upgrade comes down to proper commissioning, not luck. A good commission procedure needs to be developed between IT equipment and its supporting electrical infrastructure; it isn’t complete when UPS and PDUs are commissioned, but rather when IT equipment is installed and preliminary testing is completed. Before the cabinet is ever “lit,” all equipment should be labeled and as-built drawings completed. It can be tempting to skip this step because of tight deadlines or miscommunication between IT and facility managers, but it’s an important part of the process.
Don’t rely on house power during construction
A remarkably common oversight is failing to budget for temporary power during construction at a live data center. Many contractors assume that using house power is a given, but often find that the one-line is not as selectively coordinated as they believed, leading to delays in commissioning or testing and high-dollar extra costs.
Instead of being taken by surprise by energy needs, plan ahead for routes of delivering power from a temporary electric utility service or mobile generator — and take the time to provide a full set of contract documents that illustrate the plan for implementing temporary power onsite. A one-line or simple raiser diagram should be based on load calculations that include all temporary equipment proposed by each discipline onsite during each stage of the project.
Don’t clutter the facility with overhead pathways
The overhead space in a data center is prime real estate, meaning it can quickly become congested. In addition to your wiring pathways, overhead space needs to accommodate mechanical ductwork, plumbing, fire suppression, and structural members, so planning ahead and leaving space for future needs is critical.
The ceiling space above the data center, particularly in the electrical room, too often becomes clogged with conduit for feeder pathways. Make sure you fully utilize underground and in-slab conduit runs. By using those conduits, you can reduce collisions between trades during remodeling and save money on hangers, bending, and installation.
Include data center information management (DCIM)
The need for information about a company’s IT and facility systems continues to increase, particularly as co-location becomes more common. Decision-making in data centers needs to be supported through a dynamic, systematic information flow and analysis. If your goal is to build and maintain a data center facility for your customers with maximum flexibility and scalability to respond to every changing business need, then you need to push for a DCIM.
Not having facility information that is timely and accurate can result in negative outcomes, ranging from insufficient space, power, or cooling to excess capacity resulting in unnecessary expense.
Don’t over-simplify lockout/tagout procedures
Mission critical lockout/tagout (LOTO) procedures can’t be made simple because mission critical facilities are complex environments. Thus, LOTO procedures must be commensurate with the complexity of the space.
The objective of the power system is to provide uninterruptable power, which, by its nature, constitutes complexity and involves dual feed equipment. Every effort should be made, as Albert Einstein was reported to have said, to “Make everything as simple as possible, but not simpler.” Well-documented procedures and training are imperative.
Don’t underestimate the benefits of retrofitting an existing data center
Retrofitting an existing data center can be a cost-effective option when addressing the demands of a changing business climate. The challenge, however, is that it’s a bit like trying to remodel an airplane while it’s 40,000 feet in the air — it’s doable, but it requires great skill and planning. A critical first step to retrofitting a data center is preparing your customer for probable outages while you retrofit and ensure that your new designs are indeed flexible and scalable.
An average data center facility undergoes a design upgrade every 10 years. Some of the most common reasons for an upgrade include wanting to increase energy efficiency or redundancy and/or make capacity change in power and cooling.
What’s challenging today is that companies likely rely more heavily on it than they did when it was first built. As discussed earlier, this means you may need to use temporary generators and install new switchgear, UPS, and power distribution equipment to avoid outages.
Don’t rely on last year’s forecast for data center growth
IDC, Cisco, and Gartner are all forecasting major shifts in data centers’ computing platforms — with predictions that by 2019 only 14% of all data will be processed by traditional-type data centers. This would mean that 86% of computing workloads will be processed in cloud-type data centers in just a few years. This type of shift in the data center world is why it’s essential to design and build with future change and adaptability in mind.
McGettigan is vice president of preconstruction mission critical for Faith Technologies. He can be reached at [email protected]. Cotten, P.E., is a senior preconstruction specialist for Faith Technologies. He can be reached at [email protected].
