Custom Infrastructure Powers the SuperNAP

Switch Communications CEO Rob Roy with one of the WDMD custom cooling units at the Las Vegas SuperNAP during a tour last year.

LAS VEGAS -High on a narrow catwalk alongside the massive cooling units at the SuperNAP, a security guard stops to open one of the four doors lining the side of the unit. As the door opens, a powerful blast of air streams out. “You have to remember to hold on to keep from getting blown off,” said Melissa Young, the Executive VP of Sales Engineering at the SuperNAP, a 407,000 square foot data center facility built by Switch Communications.

The cooling unit is a WDMD - short for Wattage Density Modular Design - a custom-built unit housed outside the data center that can automatically switch between four different cooling options to deliver the most efficient cooling for current conditions. Young says the WDMDs are “built by Switch, for Switch” and not available from any vendor.

The units are part of the customized power and cooling infrastructure at the SuperNAP, where Switch also builds its own power distribution units (PDUs) and remote power panels. Young says the SuperNAP’s generators are also customized to Switch’s specifications by  Detroit Diesel.  

At a time when many large data center builders are focused on the industrialization of data center construction using standardization and bulk purchasing from vendors, Switch is charting a different path, building custom equipment to fit its vision for high-density data centers supporting power loads of 1,500 watts a square foot and beyond. It’s a philosophy also seen at Google, which builds its own servers, containers and networking gear.

The ‘Have It Your Way’ Data Center

A rendering of the Ascent Corp. CH2 data center, with dedicated customer equipment yards using different cooling infrastructure ringing the exterior of the facility.

Data center users have lots of design choices. Turn-key space or powered shell? Raised floor or slab? Chilled water or air-side economization? Most users decide on a set of desired design characteristics and search for data centers or providers that can meet all their criteria.

In its new Chicago data center project, developer Ascent Corp. has adopted a flexible approach designed to accommodate a variety of customer requirements within a single multi-tenant facility. It’s an approach to data center design that that echoes the old Burger King slogan: ”Have it Your Way.”

“We think that users are looking for a lot more choices in the ability to customize solutions,” said Phil Horstmann, the CEO of Ascent, who calls the approach Dynamic Data Center Suites.

Ascent’s CH2 site is near the huge facility the company recently built for Microsoft Corp., and will be able to leverage the extraordinary power and fiber infrastructure supporting that project. CH2 will have 50 megawatts of power capacity and its own substation. 

Plug-n-Play Data Centers, Built to Order

Digital Realty Trust has been a pioneer in building data centers faster and cheaper through the “industrialization” of data center design. The company, which operates 13 million square feet of data center space, has been able to reduce costs and streamline construction using standard components and repeatable design concepts and leveraging its buying power with suppliers.

Digital Realty (DLR) will soon be packaging its expertise in a new data center design and construction service for enterprise clients. As it enters the professional services business, Digital Realty will compete against tech titans like IBM and HP as well as data center design/build specialists.    

“We’re moving toward a plug-and-play data center with components assembled on-site,” said Chris Crosby, senior vice president of Digital Realty. “The data center will basically be a product. We truly believe the data center is a commodity.”

The new service has not been announced yet, but Crosby outlined Digital Realty’s approach last week in a presentation at the Tier1 Datacenter Transformation Summit. The new service will be spearheaded by Michael Manos, who arrives at Digital Realty this week from Microsoft, where he ran the company’s data center operations(Mike blogged about his new position this morning) .

“The data center industry is at an inflection point,” Crosby said. “Spending tens of millions of dollars building data centers doesn’t work in the current corporate environment. Most large enterprises have figured out that building data centers is not their secret sauce.”

The Million Server Data Center

What would it take for a data center to house 1 million servers? The IEEE Spectrum magazine tackled that question recently as part of a feature on mega-data centers prompted by the enormous server farms being built by Microsoft, Google and others (link via James Hamilton). The magazine offers its own vision of what a million-server data center might look like - a huge facility filled with 400 shipping containers, each packed with more than 2,400 servers and cooled by a central cooling system that pumps coolant into the rear of each bank of racks.

Is such a facility ever likely to be built? Is the hypothetical facility outlined by the IEEE the best approach to this kind of scale?

The Challenges of Data Center Virtualization

Virtualization brings the potential to deliver dramatic savings in terms of server count, footprint, power consumption and cooling requirements for data centers. For all its advantages though, virtualization also brings some unique challenges:

• Overall power consumption will be lower, but highly variable.
• There will be fewer servers, but each one will be more critical than ever.
• Applications can be dynamically reallocated at will, but the support infrastructure cannot do the same.
• The data center footprint will be smaller, but overall efficiency might still be suboptimal.

The good news is that there are practical and affordable ways to address these challenges and improve data center efficiency in the process. This paper from Eaton in the Data Center Knowledge White Paper Library looks at some of the power-related challenges and technologies that can address them.

Data Centers Move to Cut Water Waste

As data centers get larger, they are getting thirstier as well. The enormous volume of water required to cool high-density cloud computing server farms is making water management a growing priority for data center operators. A 15-megawatt data center can use up to 360,000 gallons of water a day, according to James Hamilton, a data center designer and researcher at Amazon.com.

“Water is tomorrow’s big problem,” Hamilton said. “No one talks about water. The water consumption (in data centers) is super embarrassing. It just doesn’t feel responsible. We need designs that stop using water.”

Some large data centers, like Switch Communications’ SuperNAP in Las Vegas, have implemented direct expansion cooling systems designed to use lower amounts of water. Microsoft and Google are trying new approaches that use recycled water and nearby rivers and canals to cool their massive data centers, which is influencing where these facilities are located.

Microsoft says it picked San Antonio for one of its new data centers because the local water company could provide large amounts of recycled water, meaning the project would have less impact on the city’s drinking water supply. Google’s new data center in Belgium is located next to an industrial canal for cooling, while other providers are incorporating wells and captured rain water into their cooling systems.

Role of Cooling Towers
Why worry so much about water? The move to cloud computing is concentrating enormous computing power in mega-data centers containing hundreds of thousands of servers. All the heat from those servers is managed through cooling towers, where hot waste water from the data center is cooled, with the heat being removed through evaporation. Most of the water that remains is returned to the data center cooling system, while some is drained out of the system to remove any sediment, a process known as blowdown.

When this process is played out at mega-data center scale, the amount of water required for cooling can be enormous, sometimes exceeding the capacity of local utilities.

What Would A Data Center Stack Look Like?

The networking world has always has its stack. But what about the data center? The Open Systems Interconnection Reference Model, or OSI Model, divides network architecture into seven levels. At its first conference in February, members of Data Center Pulse developed a proposal for a standardized “Data Center Stack,” segmenting data center operations into seven layers. The group’s draft proposal is illustrated in the graphic above (click the chart for a larger version).

“We all suddenly realized we need to have the same language, the same stack,” said Dean Nelson, a co-founder of Data Center Pulse. “If we can come up with common terminology, we can start to apply individual layer metrics.” That would lay the groundwork for metrics and certifications for entire data center facilities, Nelson said.

Is defining a data center stack a worthwhile idea? Can this be accomplished without an industry food fight breaking out? Share your view in the comments.

HP Sharpens Focus on Data Center Design

The barn is dead. Partitioning a huge data center into smaller spaces – whether you call them pods, zones, suites or modules – has become an important practice for creating flexible and cost-effective data centers. This trend is part of a broader shift towards data center “industrialization” - the use of standard, repeatable designs and equipment.

The largest players in the data center space are now building products and services around these concepts. Earlier this month HP announced a “multi-tiered hybrid design” approach that creates multiple zones within a facility. The new offering (PDF) leverages the expertise of HP’s Critical Facilities Services, the unit created through the acquisition of EYP Mission Critical Facilities.

A pod-based design approach is nothing new to EYP, which has been one of the industry’s leading design and engineering firms for many years. HP nonetheless is touting the multi-tier concept as a “ breakthrough approach to facility design” that will prevent enterprise clients from building wide-open barn-style data center to Tier IV standards when only a portion of the space will require that level of redundancy.