As a CFO, CEO or CIO with a data centre (or possibly several) in the organisation, users are undoubtedly aware that they devour electricity like small cities and that the up-front capital investment, cost of expansion and re-development is somewhere between painful and eye-watering.
So says Matthew Goulding, MD of Cannon Technologies, consultants, designers and turnkey builders of data centres. Here, he looks at how a very new data centre construction method is set to allow true incremental “pay-as-you-grow” whilst reducing total capex and opex.
Over recent years there have been several attempts at “pay-as-you-grow” data centre solutions – but whilst these have successfully deferred capex, they have led to a more expensive overall solution with higher lifetime capex costs and sometimes higher opex too.
There is now a new “incremental modular” pay-as-you-grow technique. So new, in fact, that data centre people may not have heard of it yet. The system was actually perfected by Cannon Technologies in conjunction with telecoms network operators around the world over many years, and has now been adapted for data centre “new-builds” and upgrades.
Reducing the power cost
Unless the data centre was built very recently, it’s probably a massive bricks-and-mortar building with cavernous data-halls full of racks or cabinets which are, in turn, full of servers and other IT “hardware”.
Each server cabinet could contain up to 20 servers – many of which actually do fairly little most of the time. So while each cabinet would only use as much electricity as two or three one-bar fires, they are very inefficient.
Hot air from all of these cabinets heats up the air (and the walls) of the great cavernous space – and massive computer-room air conditioning units (CRACs) around the perimeter then suck out the heat using almost as much energy again as the IT equipment is consuming. So users are actually paying for nearly double the amount of power their servers need.
Recent advances in server technology have massively increased efficiency. A technique known as “virtualisation” means that one piece of server hardware can pretend to be several dozen servers.
This keeps the IT guys happy because they can still have servers dedicated to specific tasks – but it means that the actual hardware works at 80% capacity rather than 10% which power-wise is far more efficient.
The new generation equipment is also much more miniaturised, so they can squeeze 300 to 500 “virtual” server cores in a single cabinet. This would allow the data centre footprint to be reduced were it not for the fact that business and user applications demand more and more servers, storage and so on year on year.
Moving to these new style servers improves energy efficiency but creates problems too. Each cabinet now uses as much energy as 30 to 60 one-bar fires and the old CRAC cooling method described earlier is far too inefficient.
So to contain and manage the cooling, a system of rooms-within-rooms has to be built (we call it cold-aisle cocooning) together with installing cooling units very closely coupled to these massive “heaters”.
Why mention all that?
Because, to take advantage of the new technologies – their increased performance and better efficiency – IT people probably need to rebuild the data centre or at least to expand it, or re-locate some of it, while they re-fit the new kit.
And whilst they will be aware of the more conventional data centre build technologies (bricks and mortar or “containerised modular”) they almost certainly may not be aware of the new “incremental modular” process which enables pay-as-you-grow and is generally significantly cheaper than conventional techniques.
So users may want to get them to do due diligence on “incremental modular”.
There are two conventional techniques, from a capex/opex perspective, against which to contrast incremental modular techniques.
Bricks and mortar
Bricks and mortar data centres have always been horrendously expensive whether new-build, re-purposing of existing building or rental. Given the need to provide for expansion, buildings have always had to be built, purchased or rented at a very considerably greater size than users currently need.
Not only is the up-front capex several hundred percent more than the current requirement for cabinet-space, but the opex costs of operating a half-empty building year on year are exorbitant too.
It can be 10 years before the building is operating effectively at 80% of utilisation. And then within another five years or less it’s full and the IT guys need another new building.
Frankly, for most organisations, it was never a sound business model. To be fair, there used to be few alternatives. But now there are.
Containerised modular is part way there
Initially hatched out of a “transportable disaster recovery” concept, equipment vendors and disaster recovery specialists started to kit-out standard ISO transportation containers as mini data centres. The idea being that when disaster struck (think flooded or fire damaged data centre) one of these container units could be rapidly driven to site and a limited IT service restored fairly quickly.
Then as many data centre bricks-and-mortar buildings hit full capacity, containers started to be used as temporary “data centre expansion” units – located side by side or even on top of each other in the car park.
It was almost a great idea. But ISO dimensions were not designed for IT equipment and for full-scale data centre use – so they have to be either half empty to allow engineer access or cabinets have to be put on strange push-pull railway arrangements because there isn’t enough room for front and back access at the same time.
To do this, flexible cabling has to be used. But IT cabling was never designed for lots of movement, so it can break with resulting IT downtime. With opportunity or lost revenue costs of potentially millions per minute.
Larger “containerised” solutions are available, but in truth they are really pre-fabricated buildings (pre-fab) and, like the ISO containers, come in fixed-size chunks. And they tend to have internal pillars which make it difficult to best utilise the space.
Also they are not easily extensible so, with both ISO and pre-fab, the “module” is to buy another one and put it next to or on top of the previous. It’s a little closer to pay-as-you-grow than bricks and mortar but both are quite expensive.
The new modular solution
For two decades or more now, telecoms network operators (both fixed line and cellular) have needed to build cheap but long-life buildings to house radio base-stations, network “points-of-presence” and a whole host of other equipment.
They also have the problem of “small now, bigger later”. Telecoms competition has forced them, in conjunction with ourselves at Cannon Technologies, to develop highly cost effective solutions based on pay-as-you-grow, but don’t pay much. The result is a truly modular building system completely optimised to accommodate IT equipment racks and cabinets.
In fact the system is so incremental that it can grow in any direction at any time by as little as one cabinet width/depth.
Cannon Technologies’ engineers developed an ingenious method for supporting the ceiling and roof structure, which completely avoids conventional pillars (which almost always disrupt the data hall layout).
Around all of this goes a specially designed insulated steel interlocking wall panel system that can cope from arctic to Saharan conditions. The whole building can be made seismically resistant for earthquake prone territories and has high snow-loading capabilities.
Unlike containerised and pre-fab options there are no major transportation issues – every element of the building is one-man hand-able. No massive trucks or cranes needed.
And not only is it a highly cost effective construction method – it is one specifically designed to give the IT guys everything they could wish for.
An added bonus is that the advanced design concepts used maximise operational efficiency, considerably reduce power consumption and reduce the carbon footprint. The pre-engineered nature of the system also gives robust levels of resilience for minimised downtime.
What users need when they need it
Users really do only have to pay for the space/capacity users need today. Extending the building is really simple – in any direction. It’s literally a case of taking down some wall panels, installing extra racks and roof sections then putting up extra wall panels. Job done, building extended.
Complex shapes are not a problem either. So if local geography dictates an “L”, “U”, “H” or “T” shape – that is very simple to achieve. This also means that sites not suitable for other building methods can readily be used.
The whole building is re-usable. So if users need to move it to a different location that’s simple too – which protects the investment.
This new technique is so revolutionary for the data centre world that CFOs and CEOs would want to evaluate it in these days of tight cash. Cannon Technologies builds every form of data centre. They have no particular axe to grind and can give impartial advice as to whether incremental modular or a variant of the existing methods would be most cost efficient for a user’s