Subscribe
About

Test-driving data centres

Therese van Wyk
By Therese van Wyk
Johannesburg, 05 Dec 2011

For a few large companies, there is a compelling case for having their own data centres. But to build, update and maintain a data centre that can take real-world knocks when needed is a huge undertaking. For starters, redundancy in energy systems, networking and IT disaster management are required to get the facility up to par.

Add to that soaring energy costs, changing infrastructure technologies, and denser and faster computing hardware that can run at higher temperatures and require less cooling. These and more make for daunting complexity; scores of technical specialists must be coordinated and managed. And then there are the multitudes of large expenses, ranging from diesel for weekly generator tests, to replacing generators every few years, to keep that data centre ready when the knocks arrive.

Faced with all this, just packing up your company's server room or data centre and moving into five-star accommodation at a service provider can look like an attractive way of ditching complexity. There'll be only one neck to wring when things go wrong. In the search, you find out soon enough that moving to a service provider may not be cheaper, but rather better value. At least now you should get the power and networking redundancy, and other big-ticket features you could not afford yourself.

Yet, letting information out of the house is a scary project, no matter how gleaming the service provider's facility. If your company has a five-star internal data centre, will a service provider really be able to deliver services at the same standard? And if your company is getting by on a few servers in a room next to the CIO's office with some air-conditioning, will the higher monthly costs really be worth it in terms of better service levels?

One way of finding out is to take a hard look at the services your company's IT department offers the business, identify some small, independent IT stacks, and run a few non-critical experiments at a promising-looking provider. If those work well, the business-critical services can follow.

Data centre experiments

Picking a service provider is a big undertaking. Sometimes, a deciding factor narrows down the field early in the process. Data centre connectivity to the outside world can be such a factor. For application provider EOH, the decision to co-locate at Teraco was driven by their customers' preferences.

"Two years ago, we were an MTN customer. Our customers were asking us to do database and application hosting for them," says Rob Godlonton, executive director of infrastructure and applications at EOH, a solution provider and Teraco co-location client.

"But in the MTN data centre, connectivity was a problem for us. Customers who had contracts with Internet Solutions or Telkom Business didn't want to change their network provider. At that time, the interconnect fee with MTN was just too expensive; it made it difficult for us to attract customers."

Without clear deciding factors, a way of sussing out a data centre provider is simply to run an experiment on a small non-critical IT stack, which includes hardware, operating system and applications. That way, it's possible to take a multitude of factors into account over a longer period. If the provider is good enough, one can migrate more business-critical services later.

"When you move from your own data centre to a hosted one, a big factor is instability in the availability of applications or infrastructure for a period," says Val Moodley, regional manager, data centre and virtualisation for Cisco SA.

"It's key to do lots and lots of testing, like migrating your development environment before migrating production. A lot of customers do that, and then work on development at the outsourced facility. This way they can test the availability, stability and functionality of the service provider."

Confidence-building

There is a lot of hype around saying everyone must move into the cloud, and if you don't you're ancient and should consider another career. It's dangerous for service providers to continue saying that.

Johan van Huyssteen, Business Connexion

But it is possible to find out even more before committing business-critical IT systems to a provider's data centre. A data centre client can actually run a series of experiments, each more business-critical than the last, to see if the provider's services are good enough. For a large company cautious with its information, this can help in moving beyond a mindset of 'I want to hug my server', prevalent among IT managers.

"There is a lot of hype around saying everyone must move into the cloud, and if you don't, you're ancient and should consider another career. It's dangerous for service providers to continue saying that, because we certainly find clients that tell us they're not ready for the cloud," says Johan van Huyssteen, GM of cloud infrastructure services at solution provider Business Connexion.

"We find that as clients build their confidence with service providers, they move away from a first stage, where they say, 'You can run this stuff in our data centres on our equipment, and service the equipment'.

"Then they say, 'We want to move our stuff into your data centres, and you manage it. In the third phase, they say, 'We're quite happy for you, the service provider, to own the equipment, but it has to be dedicated to us as a client. You can service it, own it and depreciate it.'

"And in the fourth stage, they say, 'We're quite happy now, happy to move into an infrastructure-as-a-service model, and pay for services as and when required'."

Connectivity was a problem for us. Customers who had contracts didn't want to change their network provider. At that time, the interconnect fee was just too expensive - it made it difficult for us to attract customers.

Rob Godlonton, EOH

Even if a company is comfortable with cloud computing, it will probably be concerned about new features and services offered by a provider. As an example, technology to burst from a private cloud into a public one already exists, meaning that a private cloud can make use of public cloud capacity on an ad-hoc basis.

"Even though the technology may be there, it will take time for clients to build the confidence to say, 'I'm happy to burst to somebody else's cloud at night when I need that capacity'," continues Van Huyssteen.

"One will have to combine the virtualisation technologies with security technologies, to build client confidence, make sure it is auditable, and compliant to required standards, such as SAS70. Only then will clients go for it in a big way. Large corporates and the public sector will insist on those standards. For example, if you want to do business with SARS, you have to be ISO20000-compliant.

"Clients want the benefit of everything-as-a-service, but they don't want to place their business at risk. I think that's why the uptake has been a lot slower than we anticipated. It's a lot easier to sell that service to existing clients who currently run on our data centre on our equipment, than to convince new client to come into that world.

"It really is about hand-holding, confidence-building and maturity between the service provider and the client.”

Making sure that what happens inside your provider's data centre is to your liking, as well as disaster recovery fail-over sites linked to it, is one thing. But a data centre is just part of a much bigger picture, argues Markus Schorn, CEO at Pattern Matched Technologies, an MTN Business client, and provider of value switching technology to the financial industry and credit retailers.

Ecosystem

"If you can assume MTN is the only service provider in this particular loop, then the warranty and guarantee system they offer me is excellent," argues Schorn.

"The problem is, MTN is connected to the real world via telco infrastructure like Telkom and the mobile network operators (MNO). So MTN does not own all the infrastructure that enters and exits the data centre.

“In any system, the weakest point is always the one that fails. No matter how well your data centre is managed, and even if MTN can give a 100%-compliant guarantee, you still need to look at the weakest point of failure, which is generally, unfortunately, our ageing Telkom infrastructure.

"Warranties are only as good as the complete picture. It's not just about what MTN can offer. It's about MTN and all its partners. So Telkom becomes a partner, Neotel becomes a partner, Vodacom becomes a partner when the service involves extra hops in communication. Often we'll find that MTN has done a great job, but one of the partners has let down the solution because their bit is not working,” says Schorn.

"In any hosted environment, where there is more than one service provider involved, the warranty provided by the data centre is not necessarily the reason why something fails. It is the entire ecosystem. In SA, we still have tremendous difficulties with the ecosystem because, to a large extent, fibre has not replaced the copper telecoms infrastructure yet."

When reviewing potential service providers, then, the providers' providers need some looking at. Since data in a data centre is fairly useless without a reliable, scalable way to get it where it needs to go, a data centre's networking will become more and more crucial.

Some companies really need the control, agility and scalability of their own data centres. But for most, the cost and effort in building, maintaining and keeping up with new technology can be a huge burden.

But choosing well among data centre providers is a tricky business. One way of finding out what a provider is really like is to gradually move into its data centre offering. Test-driving the provider's availability, stability and functionality over a period will help reveal its real value offering before your company locks in business-critical applications with a multi-year contract.

Share