The Quest for Rock Solid Datacentre Solutions

February 28, 2013

CamilleThe associated ecosystem of Datacentre includes network connectivity to that datacentre. So whether we are talking about Google, whether we’re talking about Indosator Cox, or a number of different providers out there, datacentre construction, this is the rationale. Underlying the content consumption, the content distribution, the video consumption, datacentre ecosystems underpin our ability to consume.

Now consumers, Ma and Pa, like your Mum and Dad, I think your question was whether your mother would understand how to configure connectivity — well, hell, no. They probably won’t. But ultimately consumers are dependent, small businesses are dependent, large businesses, service providers, our entire ecosystem, depends on datacentres.

What’s interesting is that service providers coming from that network heritage, are driving a great deal of datacentre construction. There is a boom under way in datacentre construction. Our conservative estimates are that more than 800,000 square metres of datacentre space is being constructed right now — or has been announced and is being constructed — by telecom operators right now.

Now, to try to visualise what that means, for those of you who are football-orientated, that’s 100 times Old Trafford — if you’re Manchester United fans, and if you’re not — sorry, I don’t have any cricket equivalents but I think if you were looking at Lord’s Cricket Ground it would probably be about 200 times. Or the Birds Nest, it’s 3 times the Birds Nest itself at 250,000 square metres, the biggest covered area, I think, covered building in the world.So there’s a lot of construction under way. A lot of it’s being announced. And more than half of it is happening across various countries in Asia, like in Indonesia, Hong Kong, Singapore, and not least — not least — in China. China driving a great deal of the construction and again, from telecom operators.

Now, it may take 10 years for all of the construction to actually happen in China, particularly in this city, around the city of [Hohot]. If you look at it if you wanted to pull up a map, it’s very conveniently located for other major cities and also sources of power. So there are three telecom operators investing in gigantic datacentre construction in that city alone. So, that is one of the drivers of the market.
So in terms of datacentre construction, we probably think we know why datacentres fail. There are about eight to 10 reasons why they fail. And several of them are around Acts of God: earthquake, flood, there’s fire. All of these types of issues, Act of God, number one on the list. Human error. The idiot who brings in a can of Coke into the datacentre and then spills it and causes an electrical outage. There’s a lot of that. Or similarly, the individual who doesn’t really think that temporarily citing datacentre infrastructure — and by the way, the definition of datacentre can be just a single rack in an enterprise — citing a datacentre — and this is a real photo on the bottom left, by the ladies’ facilities is probably not a good idea. This stuff happens.

But usually it’s Act of God. Top right-hand side, that’s datacentre in India, and on the bottom right-hand side that’s the picture of a datacentre in Australia where there was a flood. And I’m naming no names, but some of them were probably quite familiar in terms of what happens. So when we have issues around human error, that’s designing it properly, then there’s issue with power outage. And again lots of that is not actually the power grid not being available — sometimes that is the case– but actually not configuring the power load that you need and how that might scale. And that there are issues with equipment failure; yes, sometimes it’s a security issue, let’s not forget, people trying to hack in, trying to cause some error or cascade of error, trying to cause or just implicitly, data corruption happening, and not least any one of these issues turning into a cascading problem that drives a complete failure of the datacentre
So is it dangerous out there? Yes, it is. There is a lot of vigilance that you need to exert from the beginning — from the construction through the maintenance. It is a lifecycle of vigilance in terms of making sure that that data centre does not fail. And that’s why a lot of enterprises are considering externalising the datacentre, partly because of the complexity of maintaining their own, or if they want to maintain their own, they’ll externalise some of the backup disaster recovery to a third party. And in many cases that third party that wants to step up to the plate is a telecom operator.So one of the issues is: if you don’t think you have the best infrastructure, or you’re certainly relying on a third party for the infrastructure, how do you choose? So survivability of that datacentre against Acts of God and other issues, how do you check?

Well, there are certifications out there, and we’ll talk a little bit about that during our [date]. But not everyone certifies. And if someone is not certified, and he is just one type of certification which is from the Uptime Institute, not everyone bothers to certify on the same standards. Not all of that doesn’t necessarily mean that the construction level is bad. And by the way, certifications can mean a layer of different things. Most people — let’s take this example — certify against their design but not their actual construction. These things cost money. So you need to be vigilant to understand what is in place. There are a number of questions you need to ask.

If you are going to externalise, for certain reasons some of the datacentre functionality,you want to also have some external survivability — be very careful with what you buy. We have a growing market for cloud services, which are there to not only supplement, but in some cases supplant, physical data infrastructure that enterprises may have in place. But if you are going to externalise, is it purely cost that you should be driven by? Costs, by the way, vary enormously by configuration, by service provider, by country, by region. This is pretty anonymised data just to show different pricing out there in the market, reflecting about 40 different service providers from Google to the Indosats of the world. But you should not be driven by price. Pay peanuts: get monkeys. Yes, indeed, that may be the case. So check through the small print, and we’ll talk about what that should be.

But maybe one of the most important things that I’d like to propose to you, and this is in a sense reflecting part of what Rick was talking about, he said that with IT we’re on a journey. And I totally agree with that. We’re not only on a journey, it’s an evolutionary journey. And if we think about datacentre infrastructure in terms of evolution, and this good old Victorian, that we still think about, Mr Charles Darwin, then I think it is an important way to think about how we construct datacentre infrastructure and the datacentre ecosystem. It’s not about being big. The dinosaurs were big; that didn’t save them. It’s not about the biggest, it’s not being the most intelligent. It’s being the most adaptable to change. And moving along from the discussion about SDN, I think that’s really at the nub of what SDN is all about, and some of the infrastructure that’s being put into place. It’s not size, it’s the ability to adapt to new demands within the datacentre and also outside of the datacentre in terms of the networking environment etc. So the datacentre fabric, the external network, the systems and the processes need to adapt on a constant basis.

Camille Mendler, Principal Analyst, Informa

(Courtesy: NetEvents)

Leave a Reply