Start with cloud, then grow somewhat and take out a cage in a colocation, then graduate to a wholesale pod and, finally, reach the size and stature of needing to design, build and operate your own gigantic datacentre. That is one of the current datacentre narratives. The other is about large traditional enterprises that have done in-house datacentres forever, and over the past few years have become smarter and outsourced some or all of their datacentre capacity to commercial multitenant providers.
There are a lot of in-between scenarios as well, proving there isn’t really one way to go about it that suits everybody. Companies all have different infrastructure needs and, while various industry bodies have released countless papers about best practices, users say there aren’t practices that are really best for everyone.
The same is true when faced with the choice of outsourcing versus insourcing. Datacentre information provider DatacenterDynamics’ annual Converged conference in New York in March 2013 featured a discussion by datacentre heads of three large US institutions on this very topic.
Financial services players Morgan Stanley and NYSE Euronext and pharmaceutical company Pfizer proved there are many ways to slice the pie. All three companies use a combination of insourced and outsourced datacentre capacity (NYSE is also a provider itself), and all three are more confident about outsourcing to commercial providers today than they were a decade ago.
Ten years ago enterprises such as these were wary of entrusting the operation of their mission-critical infrastructure to commercial providers, but the outsourcing approach is a much bigger part of the discussion now.
You’re getting state-of-the-art equipment as they start to build these newer facilities and they’re built solely with the purpose of being a datacentre
As Sudhir Kalra, executive director of technology at Morgan Stanley, says: “Colo [colocation] providers these days are more sophisticated than a typical datacentre landlord was five, even ten, years ago.” Back in those days, that typical datacentre landlord was really an operator of a commercial office building who also operated datacentre space for tenants. “They just don’t have the focus that the datacentre needs,” Kalra says.
Today, the industry has multitenant datacentre buildings whose operators exist for the sole purpose of providing datacentre services, Kalra says.
And the people who work in these buildings have the same skills as people you would find in a Morgan Stanley datacentre. “Some of the people that a typical datacentre services provider hires these days are the same people I would hire if I were managing and operating the datacentre myself.” It is the same level of talent, so it makes no difference who signs their paycheck.
David Lewis, senior manager of business infrastructure services at Pfizer, says that, not only do you get professional datacentre staff from a commercial provider, you also get access to the best gear at the lowest cost.
“You’re getting state-of-the-art equipment as they start to build these newer facilities and they’re built solely with the purpose of being a datacentre,” Lewis says. Because colocation providers can leverage economies of scale with equipment suppliers and project and construction managers they can offer datacentre capacity cheaper than enterprises.
Lewis says he believes outsourcing to a colocation also provides the best operational support that money can buy, with Colo companies having access to a larger, high-quality datacentre talent pool than any individual end user would.
In his opinion, while resource challenges are the same across the board, a commercial provider can offer better career-advancement opportunities for their datacentre staff than a financial services company can. “I just don’t have that career path available for the building engineers,” Kalra says.
A piece of paper that says I’m in a Tier III or Tier IV building – when I’ve already evaluated their entire infrastructure design – is not that important
For Pfizer, this is a crucial part of evaluating providers. “We would expect the colocation provider to have more talented resources than we might have in-house,” Lewis says.
NYSE Euronext’s datacentre team has a similar philosophy. Rich Frey, its vice-president of facilities engineering, says one of the reasons why the company mixes in-house and outsourced datacentre models is access to top datacentre talent in colocation facilities.
So what are customers of this calibre concerned with when outsourcing? One of Kalra’s top concerns is sharing infrastructure with other tenants in an outsourced facility. If he has to share uninterruptible power supply (UPS) systems, cooling systems and generators with another company in the building, “that will keep me awake at night,” he says.
Others concerns surround whether engineering support is available on a 24-hour basis and when the provider performs maintenance on the mechanical and electrical infrastructure. These considerations are followed by how tight the service level agreements are. Kalra likes a good old-fashioned tour of the facility and plant room. He says just seeing the plant’s conditions and talking to operators who are removed from the property manager by three or four levels can offer up some useful information about the facility.
Enterprises also want to know that the provider’s maintenance activities do not cause service disruption. “We want to make sure that maintenance or operational tasks don’t affect our operation,” Lewis says. A 2N infrastructure is crucial, so that maintenance can be done without taking clients offline.
NYSE Euronext evaluates provider sites the same way it evaluates potential locations for its own datacentres. Frey says in the two-plus decades the company has operated datacentres, it has developed a stringent set of procedures and benchmarks to evaluate facilities against.
I’m looking for someone in the mission-critical environment that’s more like me
There are gradations, however, which depend on the application. If NYSE is looking at a datacentre site for a mission-critical application, for example, it will require the same level of resiliency and operational support it has in its own datacentres. The standards are more relaxed for applications that are less critical.
Frey says he seeks a high level of transparency on the provider’s part. Not all providers will be ready to hand over a detailed list of their escalation and change-management procedures, maintenance schedules and so on. Some of them, however, will have a hefty folder with all the above available for the potential customer. These providers score big points with Frey.
“Things like that give us comfort,” he says. “If it’s a struggle getting those, we lose transparency. I’m looking for someone in the mission-critical environment that’s more like me.”
Kalra’s team evaluates providers from a short list of companies, where they examine in great detail all the policies and procedures the providers have, including things such as operational staff and change-control policies. This information is then plugged in to a matrix and checked against cost requirements. It is a structured process that encompasses as much information about the provider as is available.
Efficiency is only one factor
Energy efficiency is part of this matrix. “PUE (power usage effectiveness) is important, but only in the context of TCO,” Kalra says.
Datacentre energy cost is high – 50% of the total operating cost in many cases – but when combined with the cost of other technology infrastructure groups, such as networks, compute and storage, the number drops to 10-15%.
For NYSE, PUE is important only if the provider charges the tenant for energy on a pay-as-you-go basis. If the rate is flat, then the efficiency of the infrastructure does not make a difference, Frey says. As part of Pfizer’s sourcing strategy, Lewis’s team will not consider a provider that does not offer pay-as-you-go energy charges.
“We want to be able to control what those utility costs are,” Lewis says. “We want to be in a region where we know we’re getting the best rate possible, so we target those areas where we can get a lower utility rate.”
Tier ratings are another aspect providers highlight in their marketing materials, even though major users may not view a tier rating as essential. Such users are sophisticated enough to evaluate the level of infrastructure resiliency themselves, rather than relying on a third-party evaluation by a certifying body such as the Uptime Institute.
“A piece of paper that says I’m in a Tier III or Tier IV building – when I’ve already evaluated their entire infrastructure design – is not that important,” Kalra says.
The bottom line is, the provider has to offer the same level of service the customer can offer themselves or better and at a lower cost. “Whether those people have a badge that reads your company name or your provider’s name is irrelevant,” Kalra says, “as long as you’re getting the service you’ve signed up for.”