How to choose the right type of colocation center

Although two of the most quoted reasons for moving infrastructure into the cloud are scalability and elasticity, another reason organizations might consider the cloud is to avoid the painfully expensive costs of building and maintaining a highly available data center. More typically, to get higher control and more customization, organizations choose to house their infrastructures in a third-party data center or colocation center that offers better protection than the company could afford on its own.

Choosing the right colo for your needs is not as straightforward as looking one up in the Yellow Pages. There are a wide array of facilities geared toward different types of users, as well as a similarly wide array of levels of redundancy and scalability. And many of the best colocation centers tend to fly under the radar for security reasons — chances are you’ve driven past them without even knowing they were there.  

Let me help you know how to find a good facility and what to look for when you visit to help you make the right decision.

The kinds of businesses that typically offer colocation space typically boil down to three categories:

• Vendor-neutral carrier hotels
• Internet service providers (ISPs)
• Internet-based businesses that find themselves with excess data center space

To add to the confusion, many ISPs and Internet-based businesses are themselves tenants in someone else’s data center space. A quick search of the Internet will yield a vast number of businesses offering colocation space — most of which are in fact leasing their data center space from a larger colocation provider or a vendor-neutral carrier hotel. Getting the full story on who you’re actually dealing with can require a lot of research.

Option 1: Vendor-neutral carrier hotels

Following the deregulation of the telecommunications industry, new competitive local exchange carriers had very limited choice of where they could host their telecom gear — generally limited to the central office of the incumbent local exchange carrier (Verizon, AT&T, and so on). Those competitive carriers thus had to lease space in a crowded facility owned and operated by their largest competitor. If they wanted to partner or cross-connect with another competitive carrier, they had to depend on the incumbent to facilitate that. Despite the best efforts of public utilities commissions around the country to ensure a level playing field, relying on the incumbent’s facilities and willingness to provide good service was a constant thorn in the side of competitive carriers.

To achieve a degree of freedom from the big incumbents, some competitive carriers started building their own facilities. However, having a single-carrier facility made it difficult to connect to other carriers to extend their reach, provide more redundancy, and deliver better-quality services. Eventually, carrier-neutral colocation facilities sprang up in which many competitive carriers could all lease space and freely cross-connect with the help of a completely neutral landlord.

Many of these same carrier-neutral carrier hotels form the highest tier of today’s available colocation space. Although largely invisible to the public, these facilities and their tenants collectively form much of the backbone of today’s Internet and telecommunications infrastructure.

Option 2: ISPs and large Internet-based businesses

Many local or regional ISPs and large Internet-based companies also offer colocation space, especially in markets where high-quality carrier hotels aren’t available. Offering colocation space is, in part, a way for them to defray the enormous cost of constructing and operating a high-quality data center for their own use. But it is also a way for ISPs to offer an “on net” location for their customers to host infrastructure that will be easily reachable from the high-bandwidth services delivered on-premise.

Unlike carrier hotels, which typically feature a veritable who’s who of global Tier 1 ISPs as tenants (and potential cross-connect opportunities), these smaller, independent colocation facilities might be attached to a small number of upstream ISPs through the network operated by the colo — making it difficult or impossible to get direct bandwidth from more than one provider. This isn’t necessarily a disqualifier, because many of these operators run extremely well-designed and redundant network infrastructures. However, it represents a single point of failure from an organizational standpoint — regardless of how redundant the network itself may be.

Option 3: Colo tiers

Beyond knowing what kind of business you’re looking at, its business focus, and actual ownership of the facility, you next want to dig into that facility’s reliability. For organizations tired of paying to operate their own on-premises data centers, that cost is not simply an issue of providing space, power, and cooling — it’s a question of doing all those things with the utmost reliability.

Building a data center doesn’t truly run up the bills until you start duplicating large, expensive components such as utility power feeds, generators, universal power supply, air conditioning, and internal power distribution to provide N+1 redundancy. If that’s what you’re looking for, you need to make sure your candidate colo providers have them — it’s by no means guaranteed they do.

Generally speaking, colocation providers are organized into a tiered system with four levels. Inconveniently, this tiering system is the opposite of the system used to rate ISPs.

Typically in the ISP world, a Tier 1 provider has global reach and interconnects with all other Tier 1 providers; a Tier 2 provider has regional reach and interconnects with some Tier 1 providers and many local Tier 2 providers; and Tier 3 interconnects with a small number of Tier 1 and Tier 2 providers (sometimes only one). Tier 1 is the highest and often most expensive tier, but with the benefit of the highest-quality bandwidth from a reachability and performance standpoint.

In the colo world, the tier system replaces the “interconnectedness” grading with generalized facilities reliability. A Tier 4 facility — typically most carrier hotels in major metros such as New York, Los Angeles, Boston, and Chicago — has redundancy implemented at every level of the infrastructure to at least an N+1 and sometimes N+2 standard. Having redundancy at the power-grid level is possible only in select areas (usually in large cities), so Tier 4 facilities are rare. However, they sport the best reliability — generally striving to achieve a 99.99 percent or higher uptime level (less than 48 minutes of downtime a year).

Tier 3, 2, and 1 colocation facilities achieve incrementally lesser degrees of facilities redundancy in varying parts of the infrastructure, often by sacrificing redundancy working back from the power grid and into HVAC and UPS. From a reliability perspective, Tier 3 facilities target an uptime of 99.98 percent (a little more than 1.5 hours a year of downtime), Tier 2 facilities target 99.75 percent (about 22 hours per year), and Tier 1 facilities shoot for 99.67 percent (a bit more than a day per year).

When you’re inside the colo

Now that you understand the different types of colo facilities, you’ll want to go inside the candidates in the tier you want to have your data center in. Next week, I’ll dig into the specifics of what to look for when you’re visiting a prospective colocation provider, covering power, HVAC, staffing, bandwidth availability, and fire suppression.

This article, “How to choose the right type of colocation center,” originally appeared at InfoWorld.com. Read more of Matt Prigge’s Information Overload blog and follow the latest developments in storage at InfoWorld.com. For the latest business technology news, follow InfoWorld.com on Twitter.