Why the Northeast Needs Advanced Data Centers And How It Can Get Them

 

By Jonathan Tombes

 

The Northeast encompasses the first, seventh, and ninth largest regional U.S. economies. It includes the global and national capital of finance, more than half of the world’s top pharmaceutical companies, 200 universities, leading medical institutions, centers of media and telecom and more. Yet its third-party data center ecosystem is lackluster. It falls short in quantity, quality and value.

The colocation data centers in these metro areas represent about 11 percent of the total in the U.S.1 That roughly equals the region’s share of U.S. population, but not its economic weight. Moreover, the size of its data centers in the core NY/NJ market is smaller on average than that of the other three major markets. (See Table 1) Less square footage means tighter constraints, but it also serves as a proxy. Smaller facilities tend to be older colocation (colo) facilities. Despite those shortcomings, Philadelphia, Northern New Jersey and Boston are among the most expensive data center locations in the US.2

table 1

Consider it another way. If the region were flush with high-capacity, full-featured and cost-effective data centers, it would win more than its fair share of large, national deals. Yet in an annual roster of the top twenty-four wholesale leases, ranked by megawatt (MW), over the past two years Northeast data centers appeared just six times.3 Companies with high-capacity requirements simply find more of what they need in the South and West.

The shortage in the Northeast poses a problem – and opportunity. Market research firm IDC has predicted that worldwide a majority of organizations will stop managing their own infrastructure over the next five years.4 Investors have begun betting that businesses will be looking for more outsource options in the Northeast. Planned additions to capacity in the region, in fact, now outpace similar activity in the top-three other expansion markets combined.5 But here’s the question: Will these efforts succeed in bringing a region rooted in the “colo-hotel” past into the advanced data-center future?

ADCs and Industry Needs

What exactly is an advanced data center (ADC)? At the vanguard are Internet companies, such as Amazon, Google or Facebook, whose data centers remain core, corporate assets. These are cutting-edge facilities, but they lie outside the third-party market.

Ambitious multi-tenant data-center (MTDC) operators have become fast followers of those leaders, if not innovators themselves. Reporters and analysts who visited Switch-owned SUPERNAP in Las Vegas back in 2011, for instance, had no trouble calling it advanced.6 The Northeast region includes some of these operators. As noted, six of the industry’s largest leases over the past two years went to regional facilities.

Not surprisingly, four of those large deals involved the financial industry. The other two were media and healthcare related. What do these and other regional industries expect from their data-center partners? Let’s consider a few of their requirements:

  • Banks and financial services companies have unique storage needs, driven by regulation (Sarbanes-Oxley). They may also require Fort Knox-like security and, for some investment firms, infrastructure to support hyper-computing.
  • Healthcare organizations have security and privacy needs, driven by other regulations (HIIPA and HITECH).
  • Biomedical and life sciences research can require intense, yet flexible compute power.
  • Telecom and cable data centers require very high availability and low latency; video servers may require high ingest and/or play-out capabilities.
  • Advertising has become a quantitatively driven industry, with data mining as critical as creative capabilities.
  • Higher education is moving varying amounts of instruction online, which in effect transforms these institutions into media providers. Some academic disciplines require supercomputing capabilities.

ADC Attributes

At another level, many businesses simply want superior performance. According to the Uptime Institute, the top three criteria IT practitioners have in mind when assessing external data centers are availability, security and connectivity.7 Here is a brief elaboration of those three attributes, along with several others that help build out the profile of an ADC:

  • Availability. Guaranteed levels of compute and storage, translated as minimal downtime, which in turn requires servers that are powered and performing tasks per a given service-level agreement (SLA).
  • Security. Defined in physical terms by levels (perimeter, entrance, floor, rack, etc.) and various processes; Location may also play a role, with obscure addresses reducing risk over a well-known and exposed site.
  • Connectivity. As critical as availability for most workloads; carrier-neutral telecom connectivity, diverse fiber feeds and options for additional network build-outs are hallmarks of an advanced facility.
  • Redundant power. Not just backup, but dedicated and customizable power connections with built-in redundancy and upgrade capacity. Yet only a fraction (13 percent) of global data centers operate at level higher than N+1.8
  • Size, plus scale. Square footage and megawatts are indicators, along with right-sized capacity and the ability to handle “web-scale” growth, with no need to re-architect infrastructure systems.
  • Efficiency. Nearly all data center operators track Power Usage Effectiveness (PUE)9 ; reaching the industry’s highest efficiencies while operating at scale requires advanced energy-efficient design.
  • Services. Beyond Data Center Information Management (DCIM), advanced data centers also offer network and application management services, including customized environments for industry-specific applications.

ADCs, Innovation and the Northeast

The question remains, how do you achieve superior performance in those areas? And if you build facilities that meet the highest standards, will businesses come? On the second question, let’s consider findings from the Uptime Institute. The factors that have led enterprises to third-party data centers are the ability to scale, lower capital cost, speed of deployment and lower operating cost. In that light, it stands to reason that building more cost-effective, right-sized and streamlined data-center solutions will accelerate adoption.

Achieving superior results is one thing for the Internet companies that design, build and operate their own massive and highly efficient data centers. Enabling MTDCs to provide businesses with comparable experience in a region already marked by high costs and limited facilities is another question.

Innovation – both sustaining and disruptive – will help. Some areas, such as PUE efficiency, involve gradual steps with diminishing returns. Whether any other innovation can truly transform the industry is an open question. Achieving the performance and cost structures needed to move this market will require a number of innovations. Here are four candidates:

Prefabricated modular units. Data-centers have typically required large investments. You buy or build a huge building, add power and cooling, and then divide it into halls or stick-built rooms or wire cages and try to sell it. An alternative approach is to acquire a facility and then use modular infrastructure components, built to the specifications of particular customers. Identified as a disruptive innovation several years ago, this technique accelerates deployments and conserves capital, which can be reflected in pricing, but it also future proofs the asset.10 Rather than being locked into a design that may have been state-of-the-art three years ago, this approach enables a data-center operator to outfit a private, containerized vault with one cooling or power environment today, and then switch it out later, as both technology and customers’ needs change.

Power density. To expand upon that last point, a related area for improvement, especially felt by data-center operators constrained by space, involves packing more power into a given area. Improved power density means more servers per rack, which generally translates into better efficiency. Virtualization, blade servers, and micro servers are some of the drivers behind higher density. Yet as density rises, so too does the need for more cooling. Containment (hot-aisle, cold-aisle) and more spread-out workloads are conventional techniques, but they go only so far, which is one reason why data centers are being built in Nordic countries and liquid cooling has arisen alongside extremely “hot” computing, such as that used in bit-coin mining. Modular infrastructure enables operators to deliver the precise amounts of power required, which also contributes to higher density.

Site location. Another answer is new thinking on where to build or locate your facility. The old model has been centered on colocation sites, the well-known, downtown addresses for telecom interconnects. Colocation is critical, but making it the anchor of a data center conceals opportunity costs. By placing a data center outside of a city, but in close proximity to fiber backbones, you can have robust links to any number of “colo hotels.” Pick that site carefully enough and you also may gain vast capacity for expansion, over-abundant power resources and fortress-like campuses that defy natural or man-made disaster.

Customized services. A final answer to this challenge involves services. As organizations continue to assess how much of their data-center infrastructure to maintain and what to outsource, they will be looking for increasingly custom-fit and nuanced solutions. As important as business continuity is, for instance, the future of multi-tenant data centers entails providing not one solution, but a broad portfolio of services. That could mean planning, deploying and managing wide area networks, supporting standard communication tools (email, phone, collaboration) and optimizing infrastructure for applications ranging from medical imaging to financial modeling to data mining to CRM and beyond.

Conclusion

Businesses once managed all compute and storage technology in house. Now they run only two-thirds of their IT workloads internally, while 25 percent goes to external data centers and the remainder to the cloud.11 That trend is accelerating. If the IDC prediction holds up, over the next five years millions of enterprises worldwide will move their servers off premises. In the economic powerhouse of the Northeast, that could amount to thousands of leases for MTDC operators. Developing this market will require a greater number of advanced data centers (ADCs). Driven by innovative design, new logistical thinking and expanded services, a rising class of smart ADCs can both reduce costs and raise standards, leading more businesses to entrust them with a growing share of their IT operations.


End notes:

1. Number of regional data centers estimated at 157, divided by total in the US, given at 1447 = 10.9 percent. Source: datacentermap.com.

2. “CBRE Identifies Most Attractive Markets for Leasing a Data Center,” CBRE, Los Angeles, Nov. 12, 2014.

3. Derived from “North American Data Centers Newsletter,” Jan 2014 and Jan 2015, Year in Review issues. The companies signing these leases were Bloomberg and Express Scripts in 2013, and JP Morgan, Bitcoin 21e6, Two Sigma and State Street in 2014.

4. “Worldwide Datacenter Census and Construction 2014-2018 Forecast: Aging Enterprise Datacenters and the Accelerating Service Provider Buildout,” Richard Villars, IDC, Oct. 2014.

5. DCKB, 451 Research, “The Colocation Market,” Dec. 2014. Posted on datacenterconsulting.com. The other three expansion markets are NoVA, Chicago and Silicon Valley.

6. See “Why, Unless You’re Running With Amazon, Google and Microsoft, You Should Never Build Another Data Center,” Kurt Marko, forbes.com, Aug. 3, 2014. In this essay, Marko reflects on his visit to SUPERNAP in 2011.

7. “2014 Data Center Industry Survey,” uptimeinstitute.com.

8. DCKB, 451 Research, ibid.

9. PUE is defined as the ratio of DC input power over IT load power. For a thorough discussion, see “Guidance for Calculation of Efficiency (PUE) in Data Centers,” White paper 158, Victor Avelar, Schneider Electric.

10. “Uptime: Flash Storage is Top Data Center Disruptor,” Rich Miller, Data Center Knowledge, May 15, 2013. In addition to prefabricated modular data centers, the Uptime Institute listed flash storage, cloud-level resilience and advanced DCIM.

11. “2014 Data Center Industry Survey,” uptimeinstitute.com.