The well-known real estate adage “location, location, location” underscores the importance of site selection but doesn’t capture the complexity of factors affecting data center placement. While many facility issues are important in colo selection, enterprises can advance their colocation selection process by narrowing their range of choices geographically.
Enterprises seeking a new colocation data center can choose among almost 2,000 U.S. facilities plus another 2,500 internationally. Data center proximity, telecom latency, business continuity, employee availability, and environmental and economic factors are key in narrowing down colocation options according to geography.
PROXIMITY TO OTHER CORPORATE DATA CENTERS
When it comes to colocation, one of the most common drivers in location selection is proximity to existing legacy data centers, especially if the company is relocating existing hardware. Many enterprises migrate across town from an older on-premises legacy data center into a nearby colo. Such “forklift moves” are less risky within the same metro area than cross-country, and existing data center staff can efficiently coordinate a local move.
TELECOM LATENCY
Network performance is a critical factor in data center location selection. The functionality of many enterprise applications degrades if the tolerable latencies for data sent between corporate data centers are exceeded. Because latency increases proportionately to fiber route distance, data transfer is much faster between nearby cities than distant ones. For example, roundtrip latency between New York and Philadelphia (80 miles) is approximately 4 milliseconds (ms), while latency between New York and Denver (1,600 miles) is around 36 ms. Enterprise application and network teams should determine any latency limitations for inclusion in the geographic selection process.
Companies operating an “active-active” data center platform to decrease their recovery point objective (RPO) in business continuity planning get much tighter RPOs when latencies are below 5 ms, which is usually only achievable within the same or nearby metro areas.
Enterprises should also verify that required carriers serve their candidate geographies if there are any questions about specific network availability.
BUSINESS CONTINUITY
For most enterprises, latency maximum distance concerns run counter to business continuity protocols. Many businesses seek new colocation placement far enough away (e.g., more than 300 miles) from paired primary data centers to minimize the risk of a single disruption event simultaneously disabling both facilities.
Site uptime includes both functionality of the data center itself (building structure, electricity, cooling, telecom, monitoring, etc.) and supporting infrastructure, such as utility feeders and telecom networks. Even a robustly designed data center that is systemically functioning will suffer crippling downtime if nearby telecom networks are aerial and a tornado disables all fiber circuits serving the building.
HAZARDS AND RISKS
Hurricanes impact utility and telecom network operations across large regions, and many enterprises reject data center placement in coastal areas from Texas to Florida. Hurricanes present not only wind damage risk to structures and utility infrastructure but also flood risk when they stall over coastal lowlands for days at a time.
Tornadoes can occur randomly across the central and southern U.S., but most tornadoes only affect a narrow property path, averaging less than 400 feet wide and 4 miles long. More damaging EF3 to EF5 tornadoes (which comprise only 6% of all tornadoes as measured on the Enhanced Fujita Scale from EF0 to EF5) have much higher wind speeds and bigger paths averaging 1,000 feet wide and 17 miles long.
Because many tornado-spawning thunderstorms in the central U.S. move along a broad path from the southwest to northeast for hundreds of miles, many enterprises place two regional data centers across a 90-degree diagonal, with one being southeast of the other, or place them several hundred miles apart along a southwest-northeast line.
Earthquakes impact buildings, utility delivery, telecom networks, and building structures, but earthquake-prone areas have been widely researched and identified on the U.S. Geological Survey (USGS) seismic risk maps. While data center buildings can be designed to resist most earthquakes, enterprises should be concerned that a major earthquake could disable regional telecom networks and utilities serving the facility for a prolonged period.
Wildfires can occur across much of the western U.S., affecting utility transmission lines serving customers near and far, but only a few colo data centers are near the wooded rural areas most prone to wildfires. Recently, some utilities have proactively curtailed electrical service to prevent fires, so users should conduct extra due diligence if considering placement in wildfire-prone areas.
Disaster and risk maps, related data, and counter-measures are widely published, and experienced third-party advisors can assist in identifying, quantifying, and ranking locations for risk mitigation.
REGULATORY ISSUES
Enterprises in a handful of highly regulated industries, like financial services or health care, choose data center locations to satisfy regulatory requirements. Project teams should discuss potential locations early in the process with internal compliance officers and applicable regulators to define mandatory requirements and any geographic limitations.
EMPLOYEE AVAILABILITY
Enterprises have traditionally relied upon skilled IT staff visiting their data centers as needed — an advantage of on-premises and colo placement over public cloud. If your application architecture benefits from on-site staff interaction, you should factor skilled workforce availability into location selection. Some enterprises seek to improve data center proximity to affordable workforce availability in cities like Dallas, Atlanta, and Denver. One lesson from the COVID-19 pandemic is the recognition that many applications don’t need IT staff proximity as much as previously believed; the “server hugger” mentality may diminish in the future.
“Enterprises seeking a new colocation data center can choose among almost 2,000 U.S. facilities plus another 2,500 internationally. Data center proximity, telecom latency, business continuity, employee availability, and environmental and economic factors are key in narrowing down colocation options according to geography.”
GOING GREEN
Since data centers consume large quantities of electricity, many enterprises are replacing older, less-efficient data centers with new colo suites to reduce their corporate environmental impact. Cool, dry climates, like Denver and Salt Lake City, offer more potential free cooling hours per year using air- or water-side economizers than hot, humid locations, like Houston or New Orleans. Cooling economization reduces the data center total electricity load, reflected in a low PUE.
Colo providers in many locations can choose 100% renewable or other electricity plans with low carbon emissions. Many states have deregulated electricity supplier choice, and while customers don’t choose their own energy mix within a colo, they can focus their search on states, like Texas, Ohio, or Illinois, with abundant renewable sourcing and favorable energy mix options.
“ComEd provides reliable electricity service to most of northern Illinois, including the data center hub of metro Chicago,” said Ed Sitar, manager of economic development for ComEd. “Businesses in our service territory have access to 100% renewable energy sources through the Illinois competitive supply market and one of the lowest statewide carbon-free generation portfolios in the country. Data center operators are able to choose an ideal supply plan from over 50 certified retail energy suppliers.”
MONEY TALKS
Economics materially affect site selection, because some locations are much more affordable than others. Land is scarce and expensive in California, New York, and New Jersey, and building construction costs are also high in those states, so most providers pass through those higher costs to customers. Colo rates are typically about 20% lower between the coasts — in markets, like Dallas, Atlanta, or Phoenix — than the highest-cost metro areas. Users can research the RSMeans Construction Cost Index to compare regional construction costs, and third-party advisors can provide market colo rates across many geographies for project planning and budgeting purposes.
Economies of scale in building new colocation data centers also reduce colo rates. In the big six U.S. data center markets of Northern Virginia; Dallas; Chicago; Phoenix; the Bay Area, and New York/New Jersey, colo providers tend to build much larger data centers than in smaller metro areas. Construction at “industrialized scale” can decrease colo occupancy costs by 15% or more compared to smaller builds.
Labor costs to operate data centers are similarly more expensive in some markets than others, usually in the urban markets where land is also expensive. Staffing salaries are lower in Kansas City than the Bay Area, and those operating expense savings are passed along to colo customers.
Utility costs also vary widely by location as a result of proximity to natural resources and local regulations. Electricity costs are generally highest in the northeastern states and California and lower in the southern, Midwestern, and mountain states, but a few markets have exceptionally low electricity prices for other reasons. Service territories offering hydroelectric generation have power costs as low as 2.5 cents per kWh, compared to 9 cents to 13 cents per kWh in some expensive coastal cities — one reason Washington and Oregon have attracted many hyperscale data centers.
ECONOMIC INCENTIVES
Economic incentives vary widely among prospective locations, so it helps to understand the differences.
Statutory incentives are available by legislative act with qualifying criteria — you’re either eligible for them, or you’re not. Negotiated incentives require project-level consent from state or local officials who approve the incentive availability and valuation. The two most significant types of incentives are sales tax waivers and property tax abatements, which can apply to real property (land and buildings) or personal property (IT hardware).
Sales tax waivers tend to be statutorily enabled, allowing an enterprise to avoid paying state sales taxes on purchases of otherwise taxable IT hardware/software and electricity. These waivers often require minimum investments between $50 million and $250 million by the colo provider. Most sales tax waivers run for seven to 15 years, and hiring thresholds are sometimes required. Some states, including Virginia, Arizona, and Illinois, allow colocation customers to share these waivers, while other states allow them only for enterprise-operated data centers. A business budgeting $5 million in annual hardware and software investments over a 10-year contract at a qualifying colocation facility could avoid $3 million in sales taxes otherwise payable at a 6% state tax rate.
“The State of Arizona passed Computer Data Center (CDC) legislation in 2013 to further encourage investment by data center colocation and enterprise operators,” said Brad Smidt, senior vice president of business development for the Greater Phoenix Economic Council. “The legislation allows for a sales tax exemption for owners, operators, and colocation tenants on qualified equipment used in the data center. With only a $50 million capital investment requirement in Maricopa County by data center operators, our program is one of the nation’s easiest for qualification. Arizona’s program allows colocation tenants to participate as well if they commit to at least 500 kW of critical power capacity.”
Another potential incentive is property tax abatement, where colo operators negotiate a partial reduction of property taxes on their buildings and pass along those tax savings to customers as lower rent. In some jurisdictions, property tax abatement is also available on personal property owned by enterprises placed in their colo suites, reducing the tax burden in those locations. A few states, like Illinois and New York, don’t charge personal property tax at all, giving them a cost advantage for this specific occupancy expense.
ADVISOR ROLE
Many factors affect location selection for colocation data centers, and most information needed for evaluation is available through online research. However, finding and quantifying the data can be time-consuming and cumbersome, so many enterprises engage advisors/brokers to assist in the procurement process. Experienced advisors rely upon decades of experience on similar projects to provide value and overall savings.
“Enterprises seeking a new colocation data center can choose among almost 2,000 U.S. facilities plus another 2,500 internationally. Data center proximity, telecom latency, business continuity, employee availability, and environmental and economic factors are key in narrowing down colocation options according to geography.”
