Business criteria for selecting a data centre

By Guy Ruddock, VP Design and Delivery at Colt.

  • Monday, 17th June 2013 Posted 11 years ago in by Phil Alsop

The first time a business decides to invest in a data centre, the discussion invariably comes down to comparing the benefits of opting for either a traditional build or a modular approach - the former offering a highly customised solution, the latter based on standardised configurations using modern production line techniques.


While a traditional build will be appropriate in some instances, increasingly a modular approach is identified as more suitable in meeting flexibility requirements and cost and efficiency objectives. So, what are the key criteria that need to be considered by your business to ensure the right option is selected before commencing a data centre project?


Space and Power Requirements
The ‘white space’ or total usable floor area required for the facility will naturally define a range of data centre options. When it comes to contiguous space, modular solutions can cater for the majority of modern data centre requirements with open halls of up to 500 sq metres that can be combined as part of multiple hall solutions. For very large, contiguous spaces a traditional build may be required.


In terms of location, modular data centres tend to have specific footprints, so traditional implementations are likely to be more suitable for awkward buildings with lots of internal pillars, tight access and low ceilings. However, if an open space is available – even, in some cases, outside the building – then a modular data centre would tend to be a better solution given the substantial investment in building modifications required by traditional implementations.


Power densities (power over a given floor area) of somewhere between 1 kW/sqm to 3 kW/sqm are typically specified for most modular and traditional data centre offerings. The consideration is really around how to future-proof for requirements arising 5 or 10 years down the line without over-specifying at the beginning. For example, you may have a business requirement in the near future for 6kW per rack but you know densities are increasing. Considering options that can be upgraded without any downtime is therefore key to real future-proofing. This allows you to invest in what you need to now and upgrade later when the business need exists, without the requirement for a new facility.


Design and Construction
Typically, a modular data centre can be built from order to delivery in anything from four to six months – whereas a traditional version would take anywhere from an absolute minimum of 9 months to two years given the level of customisation involved. So, specifying a delivery schedule of less than one year puts an organisation into the realm of a modular offering. For highly customised projects, however, a traditional approach may be the only option despite the longer delivery timescales and increased capital expenditure involved.


A particular stumbling block for traditional data centres is the tendering process, which can be very extended. This typically involves a highly challenging decision-making environment bringing together professionals from many different functions across the organisation ranging from finance to facilities, IT and real estate. For most traditional builds, the consultation process to produce a specification – with a list of potential suppliers identified and the bid process in train – could take 12 months on its own. Modular options tend to have a more complete solution, meaning this process can be reduced significantly and, in some cases, removed completely.


For modular builds, the design process is generally bounded with only a relatively small number of variations possible. This means the entire data centre, including all mechanical and electrical components, can be provided as part of a complete solution. Assuming the options are properly thought out, the product will comply with all appropriate rules and regulations. As inevitably this will not be the first time components have been assembled in that way, the design risk element is substantially reduced compared to a traditional build and this is often a significant discriminator between the options.


Project and Operational Risks
Traditional builds tend to be customised one-off designs, so naturally it is likely there will be unforeseen issues during the design and construction stage, with multiple suppliers bidding against a complex specification. Many hundreds of changes would be expected as a matter of course – all of which translate into lost time and increased cost. With a modular facility, the project is streamlined so that the complex work is undertaken off site in a factory, thereby eliminating costly changes. As modular approaches usually involve a proven, repeatable design with guarantees around delivery and quality, scope creep and spiralling costs are a thing of the past.


Operational risks have cost implications in terms of downtime and fall into two categories that need to be considered. The first is associated with maintaining and running the facility – so, for example, a proven repeatable design with tried and tested operations manuals will have less risk associated with it than one-off designs. The second relates to design risk. The final phase of data centre deployment is known as commissioning and can take anything from weeks to months. It is a given that a repeatable, proven design will result in less snags, a quicker commissioning process and a smoother handover.


So, the best way to ensure reduced design risk to a design is to build the same solution many times, test it many times and adjust the design to ensure compliance. This is not possible in a traditional build; however, it is the very essence of the modular design approach and is already built into the solution. Put simply: each data centre owner benefits from each and every deployment that went before them.


Cost Considerations
Typically, upfront capital costs are high and over-specified for traditional data centre builds. It is common to have an outlay of up to 40% of the final capital cost to provide as little as 10% of the total capacity – and this is as much as two years before the data centre is operational and any payback can begin. Modular data centres by contrast are built in a scalable way, meaning that as little as 15% of the total project cost is needed upfront and payback can begin after as little as 4 months.


Because of the high utility usages of data centres, the amount of water and energy used throughout the life of a data centre is another important factor determining total cost of ownership. Delivering a highly efficient facility can lead to significant payback over the life of the data centre. Power Usage Effectiveness or PUE is the industry standard for measuring the efficiency of a data centre. The closer to 1 the better. For example a PUE of 1.2 versus a PUE of 1.6 based on a typical 500m2 data centre can result is savings of around £3 million in power costs over 10 years. A word of caution is that high efficiency on power can be offset by low efficiency in water, so it is important to ensure that the solution delivers both.


Finally, at the end of the life of the data centre and after whatever refresh cycles have been undertaken, the exhausted facility needs to be retired. For a traditional data centre, the asset would now convert into a liability with the site needing to be cleared for its next use. However, for a modular build, components can easily be removed, leaving the original facility largely untouched and there may in fact be a residual value with reusable materials being recycled.