Apple co-founder Steve Jobs said, “Our belief was that if we kept putting great products in front of customers, they would continue to open their wallets.” Jobs thought customers would seek out products and services from companies that innovate, and that is certainly true — even in terms of colocation facilities. But, with roughly 2,000 colos just in the U.S., many of which offer innovative improvements over earlier data center generations, how do you choose?

Starting the Search

Enterprises seeking a new colocation suite will typically down-select four to six candidates for further consideration in preparation for sending them a request for proposal. To identify the most suitable colo candidates, many users begin with a simple web search referencing the desired geographical locations. Websites, like www.datacentermap.com, list the majority of colo facilities; IT consultant Gartner provides guidance to its clients on colocation providers; and many industry professionals consult fellow members of trade groups, such as AFCOM, for provider referrals. Third-party data center procurement advisors compile detailed information on candidate facilities during this phase of the project.

Narrowing the List

To shortlist the options, users generally evaluate colo candidates on 12 attributes, including capacity, reliability, efficiency, telecom, risk avoidance, managed services, financial stability, expertise, and costs. Factors most important to some enterprises, such as low cost or cloud direct-connect circuit availability, may be less important to others who might be focused on scalability or cooling density instead. Users must prioritize features and benefits when down-selecting candidate facilities for further consideration.

Capacity

One of the first qualifiers for many enterprises is capacity. Customers seeking more than 1 MW of critical power or non-shared suites with dedicated infrastructure may find only a few providers that have the available space and power deliverable on the project timeline, especially in smaller metro areas. For example, only two colos in Nashville have 1 MW installed and currently available, but 16 colos in Dallas-Fort Worth have at least that much capacity available, if not more.

Scalability

Scalability to meet future needs is important because a key advantage of the colocation delivery model is the ability to add capacity when needed without paying for it up front. For example, an enterprise might initially occupy a private suite with 100 cabinets requiring 600 kW of critical power and then add 50 more cabinets needing an additional 400 kW in two years. Smart users want facilities that can accommodate growth in both space and power over the planned occupancy period.

Many recently constructed colocation buildings include vacant shell space ready for future installation of additional data halls, while others feature adjacent land ready for building expansions. Other colo facilities feature systemic designs supporting modular increases in critical power and cooling to densify existing suites without disrupting high-reliability operations. In contrast, many older colo data centers aren’t physically or systemically expandable and can’t easily meet growth needs.

“When procuring a colocation provider, look to an organization that can empower your IT today and into tomorrow,” said Ryan Mallory, chief operating officer of colocation services at Flexential. “That means not just a spot in a data center but someone who has a network of data centers, cloud capabilities, connectivity to your customers wherever they are, and the ability to help you design a business continuity plan not only for what you need today but also provide preparedness for the unknowns to come. 

“At Flexential,” he continued, “we are expanding our data centers by adding power and additional services to our footprint to go beyond the four walls to help partner with our customers to prepare for the future.”

Reliability

Many large enterprises require extreme data center reliability, whether on-premises or in a colocation setting. Those users may limit consideration to colo facilities incorporating concurrently maintainable design principles or facilities with a Tier III certification from the Uptime Institute. Many older and smaller colo data centers do not have concurrently maintainable critical systems and would likely be deemed Tier I or II facilities if formally rated.

In vetting colo candidates, users should ask prospective providers if their designs conform to specific Uptime tiers and verify those provider claims later during proposal due diligence. Many modern colocation facilities have better systemic reliabilities and more comprehensive DCIM monitoring systems than the typical 25-year-old, on-premises data center that is being vacated, which can be noted in your project’s internal approval presentation. 

Utility Delivery

Redundant electricity utility feeders support high reliability in modern colos. Enterprises should ask providers if their data centers are fed by multiple mid-voltage utility feeders originating at different substations or dual feeders originating at different transformers within the same substation — both are superior to single-feeder designs.

Furthermore, users can ask if municipal water is delivered into the colo data center using two independently valved service lines and if the facility has a reserve water makeup tank installed.

Cooling Densities/Technologies

Enterprises embracing hyperconverged infrastructure and AI applications seek data centers for high-density deployments of 20 to 50 kW per cabinet. Many newer colo facilities can easily host 20 kW per cabinet and can cool 30 to 50 kW per cabinet using hot- or cold-aisle containment and enhanced airflow management. Most Fortune 1000 companies deploy hardware averaging only 8 to 15 kW per cabinet, which is well within design parameters for almost all modern colo facilities. Users should ask prospective providers up front about maximum cooling densities and thermal monitoring tools and then down-select candidates with appropriate cooling designs.

Efficiency and PUE

Enterprises worldwide are reducing their environmental impact and using less data center power supports that objective while reducing costs. New cooling designs in modern data centers have helped reduce PUE, which is the ratio of total power consumed at the data center to the critical power used by computing equipment. For reference, data centers built in 2005 had an average PUE of 1.5, but many modern colo facilities have PUEs ranging from 1.15 to 1.35.

Colo providers often promote a low design PUE at full occupancy, so users should ask about actual PUE achieved at lower occupancies, since many facilities operate at only 40% to 80% of total design load during early lease-up years. Many enterprises can note during internal project approvals that they are migrating out of old corporate data centers with PUEs of 1.6 to 2.2 into much more efficient colo facilities. Users can also ask providers if their data centers consume electricity generated from renewable sources.

Telecom Connectivity

Interconnection is one of the most significant drivers in colo facility selection. Users may require a specific legacy carrier or seek dark fiber strands for dedicated use connecting to other corporate data centers, limiting their colo provider options.

Many enterprises now seek direct-connect circuits to accelerate migration into public cloud platforms or decrease the latency of data transfers into cloud installations. Other users seek colocations with multi-cloud interfaces, like Megaport, or peering arrangements into international internet exchanges like DE-CIX. Users should ask prospective providers for a description of all telecom network options serving their facilities and use the information to down-select colo finalists.

“DataBank’s data center-evolved model helps enterprises with diverse or changing IT infrastructure needs design an adaptive solution,” said JP Laqueur, senior vice president of DataBank. “We operate 20 data centers in nine markets with numerous private cloud nodes and on-ramps to public cloud platforms via Megaport. This allows our customers to strategically locate workloads on whichever infrastructure platform best meets their needs: dedicated colocation for compliance or latency sensitivity, private cloud for fully managed security capabilities, or access to public cloud for scalability and burstability.”

Site Hazards/Risks

Part 2 in this series discussed area-wide risks, like earthquakes and hurricanes, that affect site selection. When down-selecting prospective colo facilities, users should also research site-specific hazards like flood zones, hazardous materials proximity, or risky base building conditions.

• Colocation data centers should be located outside 100- and 500-year flood plains, which can be researched using Federal Emergency Management Agency (FEMA) flood maps.
• To minimize hazardous material explosion risk, colos should be at least 0.2 miles away from freight railroad rights of way (ROWs), and at least 300 feet from major interstate highways — a shorter distance for highways since fuel trucks carry much smaller quantities of hazmats than rail tankers.
• Proximity to petroleum and natural gas transmission pipelines is not recommended; pipeline locations are listed on the National Pipeline Mapping System website. Calculating minimum distances for this risk are complicated since the transported materials, pipeline sizes, and line pressures vary widely.
• Colo data centers should not be near dangerous industrial sites, like chemical plants and fertilizer warehouses, and fire departments can often provide data on nearby hazardous materials storage.
• Enterprises generally avoid placement of data centers within 1 mile of refineries or within 20 miles of nuclear power plants, whose locations are listed on the Department of Energy website.
• Data centers near major airports are not preferred, and flight paths and frequencies are available from most airport noise control offices. Interestingly, three of the most concentrated data center clusters in the U.S. are located near major airports in northern Virginia, Chicago, and the Bay Area, and one data center operator even purchased “inside the fence” land from an airport board for a new data center campus.
• While rare for new colo facilities, some older data centers are located in multitenant office buildings near downtown carrier hotels. If evaluating a colo within a multistory building, users should ask providers about locations of any water and sewer lines on floors above the data center and review access to chases for fiber, power, and other utilities supporting critical systems equipment. Users can also ask about exterior wall and roof uplift wind ratings if considering a location in a tornado-prone area.

Managed Services

Many leading colocation providers have expanded their menus of managed services, adding compliance processes and tools to assist enterprises with executing a hybrid colo/cloud strategy and supplementing basic smart hands, which have been provided for years. Colo providers now offer cloud migration tools and private cloud, helpful to enterprises seeking cloud-like flexibility without losing physical control of hardware.

Because the breadth and competency in services vary widely by provider, enterprises should determine which optional services are most desirable and seek providers skilled at delivering them. Mid-sized colo providers often offer more specialized managed services than the largest hyperscale colo providers.

Provider Financial Strength

Many enterprises rate prospective colocation providers on their financial strength, since providers going out of business can be hugely disruptive to ongoing operations. Luckily, colo business failures are very rare. Six of the largest colocation providers in the U.S. (Equinix, Digital Realty, CoreSite, CyrusOne, QTS, and Switch) are publicly traded, so their quarterly financial statements are public record and can be easily analyzed.

The privately held national colocation providers are also well-funded and prepared to confidentially share documentation verifying financial strength. Most colocation providers have minimal debt, but prudent users should still confirm appropriate financial stability during the candidate selection process.

Experience and Expertise

Most colocation providers have successfully operated data centers for 10 to 20 years with impressive reliability records. Some providers have hired operations managers previously employed for 20-plus years at Fortune 500 companies, bringing valuable professional experience. Enterprises evaluating colo providers should interview the facility-level managers running the data centers under consideration. 

Customers seeking multiple sites or large, non-shared colo suites can also contact the provider’s senior leadership to get a sense of the company’s strategic direction. Evaluation of operational expertise and the cultural “fit” between customer and colo provider can affect the selection of candidate facilities.

Costs

Many enterprises narrow their list of prospective providers as a function of occupancy costs, so it can be beneficial to get preliminary pricing from providers. Occupancy rates can vary by 25% between competing providers within a single metro area — even more if comparing costs with interconnect providers in carrier hotels, which tend to be more expensive than suburban facilities. New colocation facilities built to an industrial-scale on large suburban sites tend to have lower occupancy costs in dollars per kW per month than smaller facilities.

Colo providers hesitate to provide preliminary pricing outside of a request for proposal or without a detailed discussion around the customer’s needs. Understandably, providers don’t want their quoted rates to be shared with other providers, and they seek the opportunity to personally describe ways they can offer the best overall solution to the customer.

However, enterprises can still get ballpark pricing by communicating a brief summary of their needs (required space and power, planned growth, contract duration, desired managed services, etc.) with a colo provider representative. The resulting pricing guidance may be a range of rates and will be subject to negotiation during the selection process. Third-party advisors can also provide preliminary pricing estimates and qualitative commentary about providers to assist with the down-selection process.

Research and Advisors

Researching the available colocation options to select finalist candidates can be very time-consuming, especially for project team members who are already busy with their regular jobs. Many enterprises engage advisors experienced in data center procurement to improve project due diligence, conserve the project team’s time, accelerate project execution, and prepare exhibits for internal project approval presentations.