For customers wanting to move assets out of the data centre but for whatever reason can’t shift to the cloud, a colocation, is increasingly a viable option.
Colo is where the client buys the compute, storage, and networking equipment but instead of putting it into their own data centres, they put them in the data centre of a hosting company. They still own and manage the hardware, but they don’t have responsibility for manage the facilities—heating, cooling, lighting, physical security, etc.
As such, colocation facilities attract considerable interest from enterprises. Grand View Research estimates the global data centre colocation market size was valued at US$40.31 billion in 2019 and is expected to grow at a CAGR of 12.9 per cent from 2020 to 2027.
Gartner makes the bravest prediction, saying that by 2025, 85 per cent of infrastructure strategies will integrate on-premises, colocation, cloud, and edge delivery options, compared with 20 per cent in 2020.
Colo providers in general are located in cities where they can be closer to customer sites, shortening the last-mile connection and so reducing latency. These providers include, among many others, AT&T, CenturyLink, Verizon, Digital Realty Trust, Equinix, Internap, QTS, and Coresite Realty.
There are pros and cons to both colos and on-premises data centres, according to Chris Brown, chief technical officer for the Uptime Institute. Upsides of colos include flexibility for customers to scale up and down as needed, and the cost benefits that come with the economies of scale achievable with a large, commercial data centre.
Colos also lift the burden of hiring the technicians with the skills to operate a data centre. “That is a huge item for enterprises as it is a challenge to continuously manage and find that talent,” he said.
Potential downsides include that enterprises give up direct control over the environment that houses their data, and that using a colocation facility could have an impact on whether they comply with regulations that they operate under.
Regardless, enterprise interest in colo facilities seems strong.
Hyperscale data centres like those run by providers like Google Cloud and Amazon Web Services (AWS) aren’t among the colo providers. “For starters, instead of choosing dense urban areas, hyperscalers frequently locate their data centres in remote locations where land is cheap and they have access to renewable energy like wind and hydro."
For another, colo customers have a right to physically access the hardware they’ve deployed in colos in order to manage and maintain it. The last thing Google or AWS wants is non-employees walking around in their data centres.
Lower CAPEX, no facilities responsibility
Colocation services have many upsides. They provide rack space, physical security, networking services, and cooling and power. Enterprises just provide the hardware they want to deploy there so they can focus on their businesses without having to worry about upkeep on facilities.
Hiring a colo provider enables enterprises to move some of their data centre expenses from CAPEX to OPEX, which can mean considerable savings. Bill Long, senior vice president of product management at Equinix, claimed that colo services have enabled network optimisation and implementation of distributed network architecture that saved customers 40-70 per cent of the cost of having apps served out of a central location.
“The vast majority of savings actually comes from the network costs being dramatically reduced,” Long said, so far-flung international businesses with widely dispersed branches or those with global customers can benefit most.
“Let's say you have an office in Singapore, and your headquarters data centre is in Dallas, and you're trying to you have a global MPLS network to bring that Singapore traffic back to Dallas. That's very expensive,” he said. Placing data centre resources in regional colo facilities can reduce the need for long-haul MPLS circuits.
Tony Bishop, senior vice president for platform, growth and marketing at Digital Reality Trust, echoed the 40 per cent savings. “It’s good self-funding and typically pays for itself in 12 to 18 months. That's what we see with our customers,” he said.
Scale and Redundancy
Using colocation facilities gives corporate data centres the ability to scale. “Enterprises are not standing still,” said Sean Baillie, chief marketing officer for data centre and colo provider QTS. IT needs change, which can mean having to acquire more data centre space in more diverse locations. The right colo vendors can provide both more room within an individual facility as well as a selection of sites regionally or globally.
“Whatever they buy today, is not going to be what they have in two, three, or four years, both in the building that they're in and the geographies where they decide to deploy equipment. So you have to deal with someone that's got scale,” Baillie said.
Long said many enterprises considering colos find that their applications can’t be served well out of a centralised data centre at corporate headquarters. That’s because a lot of applications require interactions among multiple clouds and networks in order to work, and colo facilities can house points of presence for multiple cloud providers.
“So you'll want to make sure that you're hosting that workload in a place where a bunch of networks are available and a bunch of clouds are available,” Baillie said. By doing so, enterprises can keep application latency within a 30-to-40 millisecond range.
At the same time, enterprises are able to locate their workloads near key pockets of users, which is another appeal of colos. Providers like Equinix and DRT have hundreds of data centres worldwide and can replicate customers’ infrastructure in multiple locations, putting apps and data closer to end users, cutting down on the distance data and application traffic has to travel, and therefore reducing latency.
Long said 85 per cent of Equinix’s colo customers deploy in multiple regions, and it is a standard practice for one of its global customer’s resources to be deployed in at least two locations per continent. “And again, the reason for that is for performance,” he said. “They need the applications to be closer to where the users are.”
He’s found that they also reap cost savings by putting applications closer to the users. Bishop said that multi-region colocation winds up being 50 per cent to 60 per cent cheaper than traditional enterprise-owned distributed data centres thanks to removal of redundant services and headcount.
That was the situation faced by Matt Toltzien, services R&D technical architect at CDW, a technology advisory and services company. Toltzein manages servers for CDW customers who use Cisco’s Hosted Collaboration Solution (HCS). Because they use very specific hardware, customers say they can't move to the cloud so CDW offers them a private-cloud service based a colo.
“I'm the service provider hosting this multi-tenant infrastructure to a lot of my customers that are running Cisco technologies on-prem today who are looking at closing data centres, looking to reduce costs, or they don't have the expertise to run the systems on-prem in some cases,” he said.
In this case, customers wanted equipment distributed over a wide area for performance. “With a lot of enterprises trusting us to host their communication systems, they wanted to make sure it's geographically separated to provide redundancy in case of natural disasters or other things impacting one specific region,” he said.
Interconnections with cloud providers
Major colocation providers can interconnect directly with multiple cloud service providers, giving colo customers the ability to access the cloud to fill a temporary need for more compute power.
Bishop said Digital Realty Trust offers burst capacity into AWS, for example, bringing AWS resources to customers’ applications or data without actually having to move the apps or data to the cloud. Because DRT connects to multiple cloud providers, its customers can share applications and data among multiple clouds via the DRT interconnections.
Long noted one specific application that lends itself well to colo: artificial intelligence. “AI is very power-intensive. Maybe your data centre is not equipped to handle that. But a colo centre is certainly built to handle AI applications,” he said.
Baillie urged looking at the physical security of the facilities. Many colo providers have a base set of physical-security requirements to get into their buildings, such as passing through a gate to access the property, then a security desk at the front door. Colos, including QTS, may also offer custom services such as full-time video security, he said.
If you decide to go the colo route, you should have a checklist to help rate the dozens of colo-provider options. said the first thing to look at is the vendor’s has a history of serving the enterprise market.
“If you're dealing with people that have turned their attention towards hyperscale and other sectors in the market over the long term, there's no history in serving enterprise, which has its own unique set of requirements,” he said.
Toltzien also advised potential customers to do colo site visits. “I wanted to go in there to see security procedures and evaluate that for myself, you know, look at infrastructure, make sure things looked solid and fairly recent and well maintained,” he said.