The Ultimate Guide to Colocation Pricing
The former Head of Data Centers at Uber walks through the factors that impact colocation pricing and a real world example on how colocation is priced.
Cross Connections / ICX
Feb 3, 2022
Colocation is the practice of housing a company’s computer infrastructure in an offsite, third party data center. With colocation, your third party partner provides access to cabinets and/or secure cage space as well as fully redundant electrical power, cooling, and a variety of telecom carriers and internet services providers - typically for a set monthly fee.
Because of these benefits, the global colocation market was valued at $44.42 billion in 2020 and is expected to grow by a Compound Annual Growth Rate (CAGR) of 13.3% from 2021 to 2028.
But colocation is far from “one size fits all”. Every company has unique needs that impact your colo requirements, which makes understanding the cost of your colocation even more complicated.
This post walks through all of the factors that impact colocation pricing, shares a real world example on how colocation is priced, and shares data on average cross connection costs we’ve seen in the market.
About the Author
Jim Weinheimer is currently the Director of Solutions Engineering at Virtual Power Systems, leading the solutions engineering team in architecting and deploying software defined power solutions to unlock stranded power capacity in data centers around the world. Jim formerly served as the Head of Data Centers at Uber Technologies, Inc, responsible for the design, construction, and operation of Uber's core production data centers. Jim also serves as an advisor to Lightyear.
Factors that Impact Colocation Pricing
Below is a comprehensive list of the factors that impact colocation pricing. I’ve broken down these factors as they pertain to:
1. The colocation facilities themselves
2. The underlying network that connects the data center to the Internet/Cloud/Offices, etc.
Colocation Pricing: Facilities
Live photo of Jim working the racks
The“Design Point” for Power & Space
Typically, a data center has a designated “design point” as the standardized unit available for lease. When planning and constructing the data center, providers will match space and power/cooling capacity to that specific design point within the facility (i.e. “7.5 kW per rack”).
This means that straying from that design point in either direction will incur higher unit costs; using more space per rack means they can't build out (and therefore lease) as much power capacity, while using less space per rack means they have to provide more cooling in the same suite space (and they’ll quickly run into critical design constraints for the HVAC system in this direction).
The design point I've seen most frequently is 30 square feet per rack.
Next I’ll discuss power and space as separate pricing factors to help you better gauge the breakdown.
With colocation, your monthly fee is impacted by the cost to power the colocation center, ranging from the cost to run electricity to the IT equipment, to powering the HVAC to keep the servers cool, to keeping the lights in the building on.
As you’d expect, it’s hard to estimate the amount of power your colocation environment will consume; the design efficiency (and “design point”) of each data center will vary, cooling requirements will change with the seasons (unless you’re colocating in the Arctic, perhaps) and the cost to simply keep the lights on depends on the market you’re in.
One thing you can do to get a better understanding of your colocation power costs is to find out how efficient the power usage at your data center is. This is measured by a widely accepted metric called Power Use Effectiveness (PUE). The closer the PUE is to 1.0, the more efficient your data center is. But given the horrible air flow for shared colocation spaces and different machines and cabinet densities, operators are lucky if they run a PUE below 1.6.
PUE is typically factored into your colocation cost as a “multiplier”, so the lower your PUE the lower your pricing (see the pricing section later on in this post). When procuring colocation, you should negotiate a PUE “cap” that your multiplier cannot exceed; most enterprise colocation environments are running a PUE of ~1.45 and capping it at ~1.55 in contracts. If you are taking a hyperscale sized commitment, your PUE could be closer to ~1.3.
With colocation, you are able to rent out a custom amount of data center space from your provider. All else being equal, the more space you want, the more expensive your colo will cost.
To put it simply, colocation space is measured in “units”, racks, cabinets, and suites. One Unit (or “1U”) is equal to 1.75 inches in height of hardware and a typical colocation “rack” is 42RU - 48RU tall. Retail colo space is typically quoted “per RU” (RU for Rack Unit) or “per Rack”, but you can also buy colocation solutions on a “per cabinet” or “per suite” basis when procuring wholesale colocation space. Utilizing at least a full rack is most common with colocation.
This article provides a solid run down on how to estimate your colocation space requirements.
Location, Location, Location
The location of your data center can impact the cost of colocation in more ways than one.
If you want to be in a top tier, high demand market such as LA or NYC, you will pay a premium.
On the flip side, if you want to colocate in a remote location where there’s limited provider options, you’ll also likely pay a premium due to the supply/demand imbalance.
Other items to consider when examining your colocation partner’s location are if they are in a geography that commonly faces adverse weather events (I.e. are they near a fault line?) or security threats (i.e. are they in an area that is high-risk for terrorism?).
It’s important to have an adequate amount of redundancy in your colocation environment so that your systems can stay up and running even in the event of any sort of outage. Depending on your needs, you can configure your colocation environment to run with several layers of redundancy across power, connectivity, appliances, cross connections and more.
“Tier 4” data centers boast the strongest redundancy and uptime guarantee, while “Tier 1” risks it with less power redundancy. Uptime Institute’s definitions are the widely accepted definition here.
As you’d expect, the more redundancy you require in your data center, the higher your costs will be.
Security and Compliance
Businesses operate with different levels of security requirements and compliance regulations.
For example, Financial and Healthcare companies typically harbor tons of valuable information which requires them to have a number of security checks in place in order to comply with government and industry regulations.
Security and compliance capabilities and standards will vary across colocation providers. The more stringent requirements you have, the more expensive your colocation will cost.
Colocation Pricing: Network
Colocation would be useless without the ability to transmit data in and out of the environment. Enter internet connectivity!
You should be able to procure best effort broadband, dedicated internet access (DIA), or IP transit in any colocation environment, but dedicated internet access and IP transit are always the preferred methods of connectivity for a mission critical use case such as colocation.
People sometimes use the terms dedicated internet access (DIA) and IP Transit interchangeably; here I outline the similarities/differences between the two:
DIA and IP Transit: Dedicated all Day
The key similarity between DIA and IP transit is that they are both dedicated internet services.
With DIA and IP transit, the bandwidth is all yours (no sharing with your neighbors), symmetric (same upload and download speeds), and guaranteed by the Internet Service Provider (ISP). Additionally, a dedicated internet connection comes with a Service Level Agreement (SLA) that guarantees the uptime of your network >99.9% of the time, or your money back.
These are must-have internet connectivity features in a colocation use case. If you choose to utilize broadband for your colocation connectivity, all of your data transmission is at the mercy of the public internet, and subject to significantly more jitter, latency, and packet loss.
IP Transit: Peering Included
IP Transit is the term used when referring to dedicated internet connectivity that comes with a well peered network included. This means higher quality connectivity at a (likely) more optimal cost for you. The higher connectivity quality is due to the flexible routing that a well peered network provides (i.e. if one of the providers in the data center has an outage you can route over another provider’s backbone because you are already peered with them).
The (likely) optimal cost is due to the high supply of ISPs in a typical data center. Pricing is lower as a function of supply/demand, especially when you compare it to the cost of paying to peer with each of those ISPs separately.
That said, even in a data center, Dedicated Internet and IP Transit will be more expensive than best effort internet. Lightyear covered the topic in this Guide to Dedicated Internet Pricing.
Bandwidth is the amount of data that your network can transmit in a given unit of time. Bandwidth is a resource that the telecom provider owns and has built infrastructure to support, so the more bandwidth you are taking up the more it will cost you.
With traditional dedicated internet access circuits, you usually have the choice between fixed bandwidth and burstable bandwidth. In data centers, however, providers have moved away from the burstable bandwidth model over the last few years. Nowadays you’re required to fall within a circuit “commit to capacity” ratio, typically ~20%. For example, if you have a 10Gbps port in the colocation facility, you must have a minimum commit of 2Gbps.
ICX / Cross Connections
A Cross Connection (aka cross connect, XC, or ICX) is a cabled connection between two systems within a data center. Cages, server racks, closets or entire suites can be connected with direct cross connects that forego utilizing the Meet-Me-Room (MMR) fabric.
Depending on your network and application peering needs/desires, you’ll likely be procuring some (or many) cross connections in your colocation environment. These come at an additional monthly cost which I discuss further in the pricing section below.
Location, Location, Latency
The location of your colocation facility can also have an impact on your network latency, so we’re including location in this section as well.
The farther distance your data has to travel to reach your colocation facility will result in increased network latency. While this might be miniscule and perhaps not even visible to the human eye, employee time lost to latency can add up and lead to productivity losses for your company.
So when choosing a colocation facility, if the cheapest option is also the farthest away from your offices/warehouses, the cost/benefit of choosing a closer facility might make sense.
Colocation Pricing: Other
We’re including these colocation pricing factors as “other” given they are often handled in-house (but not always), or billed separately from your colocation costs.
Level of Support/Services
Enterprises can select from a wide range of support levels when evaluating colocation providers.
In some cases, the provider supplies barebone options, a location and power; the client places and manages their equipment. In other instances, the vendor acts as the business’ de facto IT team, installing the systems, monitoring performance, and solving any problems that occur.
All else equal, the fewer the number of services provided, the less expensive the service.
Despite a company’s best efforts, disasters can knock your data center offline and cause serious harm to your business.The definition of “disaster” here includes natural disasters (i.e. a hurricane) and (wo)man made disasters (i.e. a cyberattack). That’s why all (smart) companies have a disaster recovery plan in place.
Old school disaster recovery might include running two separate colocation environments simultaneously, but as you can imagine this is an inefficient and costly process. More recently, companies have been outsourcing their disaster recovery to flexible and turn-key Disaster Recovery as a Service (DRaaS) providers.
The level of sophistication with these services also varies along with the price.
Colocation Pricing - Real World Example
The example below walks through how to calculate the monthly cost of colocation.
Chart Source: Lightyear
See the below sections, labeled by letter, as they pertain to the chart above.
A - kW of Capacity
This example is for 25kW of colocation power which equates to ~5 racks based on standard rack density units (~5kW per rack). Reminder that this will vary based on your data center’s design point.
Assuming each rack is 42RU and each server is 2RU, it would yield 100 servers and 2 top-of-rack switches. Going a level deeper, if each server is dual socket (two CPUs) with 22 physical cores each, that's 4,400 physical cores of compute power.
B - Average Power Draw
The average power draw of 60% is an assumption based on the typical power utilization that is seen in any given colocation environment. When configuring a colocation space, your IT team will usually procure the “realistic maximum” amount of power draw your environment will need (based on your applications, future needs, etc.), but in reality, colo environments typically use 60-65% of that maximum at any given time (and you are only billed on what you actually utilize).
So what happens to the 40%-35% of underutilized power? That’s actually what I’m working on over at Virtual Power Systems…
C - Power Rate
This is your standard cost per kilowatt hour which will vary based on your geography.
D - PUE
As for PUE, we discussed that in the pricing factor section, and won’t rehash that here.
F - Price per ICX / Cross Connect
The chart below showcases average, regional cross connection or “ICX” pricing that Lightyear has observed. Your pricing will vary not only by the region, but also by your colo provider and the peering environment within the given data center.
Chart Source: Lightyear
As you can see in the chart above, Australia and Europe boast significantly lower per ICX costs while South America is on the higher end of the cost spectrum due to the supply/demand imbalance in this market.
K - IP Transit Monthly Cost
For more info on this section, check out the Ultimate Guide to Dedicated Internet Access Pricing.
If this guide hasn’t made it clear enough - pricing out a colocation solution is an elaborate exercise with many different factors at play.
If this sounds like something you’d rather not spend your time on, check out Lightyear’s 2-minute online questionnaire where you can procure your colocation and networking needs entirely online.
IT Purchase Process
Point to Point
WAN Connectivity Pricing Guide: P2P, MPLS, Dark Fiber & More
This report covers how WAN services (P2P, Wave, Dark Fiber, MPLS & SD-WAN) are priced including proprietary pricing insights from 15k+ telecom service quotes.
Not ready to buy?
Stay up to date on our product, straight to your inbox every month.