The measure of a data centre

CRN You mentioned scalability and how important it is these days. Are more and more businesses demanding scalability in the data centre, particularly as more business is done online and you’ve obviously particular verticals, for instance hospitality or events management, where they have a need to burst and that sort of thing, have you seen that?

Paul Yes, absolutely. It’s something we noticed a while ago when we introduced our technologies to the market with our containment modular systems, and you know people were fed up with deploying at huge cost data centres that they expect to last for 15 years with massive upfront cost, utilising very little of that capacity from day one, highly inefficient from a CAPEX and OPEX perspective, and then we also saw peaks in demand, the market going up and down and the companies were left with some massive expense on their books, so that ability to scale up and redeploy is something we’re seeing customers pushing for.

Greg It’s all pervasive now. There’s government tenders out now that say ‘describe to us how you will support 3 or 5 kilowatts per rack today and how will you, in the same footprint, support up to 25 kilowatt racks in the future’. But they only want to pay for the 3 or 5 kilowatts today. 

Now, unlike IBM with mainframes and the golden screwdriver, you know turning on more capacity, or being able to burst through a broadband type of arrangement as your internet requirements increase from time to time, it’s very, very difficult to align that type of burst capability, or that type of growth, integrate that into a data centre where power and cooling is a lot more concrete or set in stone than those other forms of IT. 

But it’s all coming together and getting more and more demanding in that way.  So if people can’t predict what IT they’re going to have in a year’s time or in three years’ time, then the best thing that I like to say to scare the pants off customers when they’re kicking tyres around data centres is if the IBM and the HP rep in your city today can’t tell you the engineering specifications of next year’s device, well how can you do a data centre that’s going to fit the purpose in three years, five years, ten years, let alone fifteen or twenty years, when you need a data centre to be around for fifteen years to get a return on investment?

Because it’s a real estate investment. That’s been a huge differentiator for us, because we have that flexibility built into our design, but it’s the business model that matters and the infrastructure underneath that facilitates the business model.

It’s all about how future proof can you be and how modular? Can you adopt changing technology over time in the data centre? Clients are becoming more savvy, but they also need to understand that they need to be efficient over the full life cycle of the facility and normally facilities have been incredibly poor for the first 10 years while you fill them up and then they’re horrendous for the last 10 years, because it’s all old infrastructure and there’s nowhere to grow.

So there has to be better ways and there are better ways of doing it, and so it’s a good differentiator certainly for us.

Marcus There’s changes in the market that resellers and consultants have to take into account. The use of the word ‘cloud’ you can summarise as virtualisation. Virtualisation technologies have been around for quite a few years right? The interesting thing now is where you’re putting multiple computer power in, multiple operating systems on the same computer. 

The predictability of the load now is actually getting less and less tangible. Take for example ‘The Voice’ on Channel 9. You have peak loads, predictably on a Monday night. But suddenly they do a Sunday show. They dropped in there outside of schedule, and where you’ve been planning this once-a-week cycle, it’s now a twice-a-week cycle, and you can do the same thing around any of the things that have an event. Take tax returns for the ATO. 

All these different agencies will have different life cycles of the use of their applications. Also if you go inside the stock broking arena, where they’re actually looking at high frequency trading, their life cycle is actually around the process of power, and they’ll change processor and therefore computer platform when the next one comes out, such as the latest release of Ivy Bridge.

They can chuck in a processor and stick in a data centre, they’ll do it, because the volume of transactions that they’re doing is directly dependent on the computer results.  So if you’re running your own environment, you can be a little bit more predictable, because you can understand your business and therefore understand ‘what’s my technology refresh?’

The challenge is when you go to an outsourcer. If they’ve not taken into account that their market at one end could be traffic light system, right, which has got a very slow life-cycle, and the computer resource is predictable, through to the opposite end which is an entertainment business that might only need your resources for an hour a week, where they’re doing an on demand say a TV show. 

If you can actually create a business that shows those different types of metrics and how you’re going to do the lifecycle, and I’m emphasising what Greg was saying, if you have an architect to put this unified architecture together to allow you to upgrade your systems during the normal lifecycle.

If you actually look at creating your pathways in your architecture, so you’ve got pathway A and B, and you’re planning for A1 and B1, so you can actually do these in-service upgrades, whether it’s computer, cabling, battery capacity, storage capacity, physical rack space – if you can create that approach that gives you modularity based around what the predicted lifecycle is, you’re actually on a winning strategy, because you’re now flexible. 

If you’re constrained by your physical environment that you’ve got all this stuff in there and you suddenly have got to go from CAT 5 to 6A to CAT 7 and you’re going from 1 gig to 40 gig and you have to shoot the place down to re-cable it, you are out of business before you even make the announcement, because people won’t allow you to shut down the computer platform – whether that’s an internal business unit or if you’re an outsourcer, if you haven’t got that prediction and agility or flexibility there, go and do something else, go and sell vegetables down the road.

Trevor Resellers aren’t actually getting the customers to go back in time and show the value they’ve delivered. So, for instance, if you go in there and convince them to change their platform to enable them to scale up over time, you’ve constantly got to go back and remind them about the strategy that they’ve got, because they tend to lose their way.  Why I say that is if there are big savings to be made, and they do need to make a small investment to scale out, you’ve got to remind them the savings they’ve had to really free up budget to make a decision to move forward.  When they go to a selected outsourcer, maybe a CSC whatever, they’re factoring all those upgrades in the background,, whereas internally they still have to justify it.

One of the things resellers tend to constantly keep on selling the benefits that they deliver to the business and enable them to move to the future. So they will want to have sales people with mobility, that’s all great, but how are they going to access it and where’s it going to run? It’s going to run a data centre. Start planning for upgrading the data centre now and the computer power as well as the power manager, because it’s going to put much more load back into the data centre. They paid for that and it’s savings, so it’s not increased budgeting.

Andrew That’s a good point. For me, it’s about getting resellers to know how they can help their customer transform their business. It could be something as simple as identifying stranded  power capacity in their data centre, so that they don‘t have to go out and spend lots of money on refreshing the data centre infrastructure technology. They can just help them identify and that makes them become more agile, so that they can react to changes in their business, in this environment.

Jacques A few years ago I was involved with a reseller working with SBS the Tour de France. SBS  told me  they were planning to create some live broadcasts and last night I’m watching the Tour de France on my iPhone. My wife’s sleeping and she wakes up and says ‘what are you doing?’ watching the Tour de France, and I’m just saying that because SBS had a challenge three years ago to design a facility that could allow them to do this broadcasting of the World Cup and the Tour de France on smartphones. In the end it’s what you can see on your apps and how you communicate.

Jason That’s an excellent point Jacques, and people often ask what is actually driving the data centre business and there’s a couple of things. One of the main things is smart devices.  So the amount of smart devices that have permeated the market like your iPhone, your androids, your tablets, that type of stuff. It was the advent if you like of the smart type devices which has driven that data centre demand, because that information has to be created, transmitted and stored somewhere. IDC reported that by 2015, 91 percent of internet data would be video.

So how do we deal with that? That again is in the data centre site. So there’s three things that I like to talk about when we talk about the data centre market and what’s driving it. I call it the three Cs. It’s consumption, which is the smart devices, compliance and cloud.

Now on the compliance side, you’ve got quite a few things coming in. We’ve talked about CFOs and CTOs and their roles and so on, but one of the things that is happening around Australia is obviously the price of energy is increasing, carbon tax and so on, but one of the things that people don’t realise is that power is actually at its cheapest on a wholesale rate than it’s ever been. So the only way for power now is up.

What’s high at the moment is the network infrastructure charges, because the providers had to put in the infrastructure for peak demand. Now with the introduction of carbon tax the price of energy is going to increase exponentially.

One of the other things we’ve seen, and resellers have got to get their heads around this, is that there’s a rating system that has come out of NSW called NABERS (National Australian Built Environmental Ratings Scheme). It uses a green star energy style rating system and is now being rolled out around Australia.

One of the things they’ve found out is that they’ve rated a building using NABERS, but as clients have brought in their computer rooms or data centres, it’s actually thrown everything out.

It’s impacted on that owner of the building being able to lease more space, because their NABERS rating has been reduced. That’s happened so dramatically that NABERS has had to bring out a rating tool for data centres.

Energy efficiency is now being assessed at data centres around the country to get the baselines we’ve been talking about.

We talked earlier about skill sets, in particular the skill sets resellers need.

So generally when we talked about racks, we talked about cabling, we talked about power – it’s not the traditional IT reseller’s market. I remember being in reseller land years ago and selling servers and networks and stuff like that.

When it came to the racks and cabling and that sort of stuff, we would go out to a cabling contractor, or we’d go out to some other place. So now that IT needs to be selling more of this traditional mechanical or facilities type product there is a skills gap and they need to educate themselves on in order to achieve best practise.

They need to educate themselves on what the actual compliance is and what the standards are around data centres. There’s a lot of standards around but if you look at the data centre code of conduct, it’s actually brought all of those standards together into a fairly easy to read common sense document.

So in fact it’s become the norm in Europe. They’ve looked at in Japan and the US Department of Energy have looked at it and now NABERS in Australia is looking at it.

So one of the tips I give to resellers is to familiarise themselves with that code of conduct and best practice, because quite often when we go out and talk to customers they ask us what best practise means.

People really didn’t have anything to refer to. For example ASHREY (American society of heating refrigeration air conditioning engineers): they’ve had a rating for cooling for around 22 degrees, 50 percent humidity around data centres, and mainframe rooms for many years.

They’ve actually expanded that system out now so that you’ve got higher thresholds for temperatures and humidity that you can actually run equipment in to become energy efficient.

Marcus Let’s come back to what’s driving the change in use. If you started with your typical comms room, you’d have say in a medium sized office in the city – a lot of people are saying ‘I want to reduce the total power that IT has’.

In a typical data centre or a typical comms room, the computer resources may be 1.2 kilowatts a square metre if that. With the move to BYOD for instance you’re now moving potentially to about 3 kilowatts a square metre inside that local comms room inside that office space – because there isn’t the bandwidth to do virtual desktop (in Australia) all the way back into a centralised data centre.  It will happen in certain instances, but typically they’re going to consolidate 30, maybe 40 PCs per RU in a local comms room – and you’re now going to suddenly jump into a power environment. 

You are going to reduce the overall power that the building’s going to take for IT, but what you’ve done is you’ve consolidated it into one point. So if you think about computer resources which are centralised and then distributed – what you’re doing now is the same thing with power, from centralised on the old main frames to distributed onto a desktop.

You’re now bringing the power back. With BYOD, it’s going to draw somewhere between 15 and 30 watts.

Marcus You have now centralised your power requirement, concentrated your power requirement in the one point. There’s some improvement, but you’ve reached that bell curve of physics, of how much energy does it take to do one computer site. 

You’re not going to see this two or threefold increase that we’ve had of efficiency. It’s now talking single digit increase on each level of processor, and like Jacques said earlier on, there’s an IBM blade server coming out that’s going to take 65 degree Celsius on the output and they’ve got to cool it.

Now where do you dump that heat? You need to have an efficient way of taking that heat and transferring it back and getting rid of it.

The rate of change you can do on your software now is actually increased because you’ve got these central SOAs  and now it’s just a client. It’s not hardware dependent and  the flipside is creating these local hotspots that the comms room is designed to put a PBX in, print server, and possibly file storage, maximum 50 kilowatts, and potentially now in some of the office environments 500, 600 kilowatts to run a thousand virtual desktops out of, and then when you actually take that back into the core data centre, the growth in data and that mega type data environment now, one of the biggest power tools isn’t the servers any more, it’s the storage arrangements. And that ain’t getting any smaller.

You can only have so much of this efficiency per centimetre of disk space, as you grow and grow that storage environment, data never decreases right.

Everyone wants that legacy data and the mining of that data now creates more and more – so you’ve got these layers of applications.

So yes really efficient enterprise data centres are important, but everyone’s forgetting it’s middle range, which is how do we actually create that efficiency.

Again bring it back to metrics and compare against the industry. When you’re doing your planning look at someone who’s done a virtual desktop for one thousand PCs in that office environment.

Trevor We go into a lot of data centres at the moment where everyone’s got all the bells and whistles, they’ve bought the most up to date. But they haven’t trained their people on how to actually configure them. 

So you might have the capacity to go down and cap the amount of power consumption in a server, but the reality is that most customers take things out of a box, and they kind of realise there has got to be power cables to plug it in.  

They’ll rack it, there’s an internet cable, they’ll put it in, and then they’ll put it in and say oh suddenly someone recognises there’s something wrong.

So there are some good opportunities for resellers to go into existing environments and try and make them a lot more efficient to optimise them, and it is a consulting engagement. It doesn’t include selling hardware often, although it could include selling appliances – like with APC in terms of some of the management of the environment. But really a lot of customers don’t do themselves justice. They buy these servers and then don’t use all the tools.

Greg That’s very interesting. It’s almost like a diminishing scale. When you think about the smaller organisations that have a number of racks, like six racks, something like that, maybe more – it actually costs a lot of money to actually employ all the whiz bang technologies, and it costs a lot of money to have the skills on hand to make full use of the technology.

Sometimes it’s easy just to use a dumb approach that is what you’ve done in the past. Whereas as you go up the major computing organisations, banks and financial institutions, all the metrics that Marcus mentioned around costs, all the big banks and things like that, they’re smashing all those things and they’re competing, and you know Commbank has got a significant jump in customer satisfaction through investment in IT which has enhanced the customer experience. So they are competing on the IT front.

But the smaller you are, the more challenging that is, relative to the big end of town.

Greg Exactly, and that’s where there’s opportunities for resellers.