Expanding a data centre

Demand for data centre space continues to defy gravity, rising despite a weak international market and relatively slow economic growth in Australia.

Virtual desktop environments are a leading factor in data centres needing to expand. With apps, security and operating systems all stored in the cloud rather than on individual devices, data centres must be bigger, faster and have more capacity to cope. Australia is a mature market for virtual workspaces, with an above-average uptake among enterprise companies.

Ward Nash, VP for ANZ at IGEL Technology, said, “A rise in virtual workspaces is leading to higher demand for data centre space. Australia is a mature market for virtual infrastructure, which is leading investment towards the country. For example, at IGEL we provide thin clients and a range of operating software designed for virtual desktop infrastructure, and we recently moved our APAC HQ from Singapore to Sydney to take advantage of the opportunities here. This is also leading to many data centre facilities looking at expansion as a means of coping with demand.”

Research conducted by Digital Realty Trust back in 2010 predicted that 80% of data companies would look at expanding over the next few years, a statistic now backed by Frost & Sullivan reporting that data centre services revenue in 2014 was $872 million and a report by Paul Budde putting total data centre investment for the last five years at approximately $5 billion.

With growth expected to ease over 2016 and 2017, as fewer new data centres enter the market, expansion of existing facilities will play a key role in maintaining a strong data market in Australia.

Cloudera and Intel have just released results from research (conducted by Unisphere) whereby 319 business leaders directly responsible for data centre operations were asked about their use of big data and what factors would contribute to the evolution of data centres over the next few years.

The research showed that IT plays the lead role in initiating data modernisation projects and that major concerns include data governance and security, as well as accessing data stuck in silos.

Commenting on the research, Sean Anderson of Cloudera said, “Accessing data stuck in silos is the most challenging element in creating an impactful data pool for analytics.

“Walling off data and limiting access is great for security, there is no other sure-fire way to keep data safe than creating real physical separation. But this has its drawbacks. Limited data access means limited insight. How do we expect our analysts and data scientists to build the best models on limited data?

“Not only is this cumbersome for analysts but it creates operational complexity and lower ROI due to managing multiple systems. So for the data management professional, it is not surprising that 38% of respondents viewed data being stuck in data silos as a pervasive issue.”

Twenty-eight percent of respondents to the Cloudera/Intel survey listed “network saturation” as a major challenge presented by the growth of data volumes associated with big data analytics, and more than half mentioned the performance of I/O and analytical applications as being a direct challenge.

This is a reflection of the challenging and multifaceted nature of a data centre build. To maintain performance and security while expanding the data centre in a careful and structured manner that will bear out sensible ROI, without overstepping the mark, is a delicate balancing act. To expand, or to create a new space — that is the question.

Looking at the equation from a return-on-investment perspective, the five most common yet challenging considerations are:

1. Whether to build a brand new facility or whether to convert an existing site into a data centre.
2. The length of time that it takes to source a prospective data centre site.
3. Whether or not the site selected has the power capability required for a data centre.
4. Risk assessing the surrounding environment to consider potential issues such as flood areas that can cause damage to the facility.
5. The time and capital required to acquire and develop a data centre site or building. Projects need to manage factors including structural and legal reviews before a site can be signed off for use as a data centre. Without effective planning, many projects run off course or are put on hold.

Based on these factors, there is significant sense in expanding a data centre, rather than risking the capital investment required to build a complete new space.

Optimising an existing facility to cope with increased demand is also not without issues, but does mitigate many of the above ‘pain points’. Of primary importance is a very thorough analysis of the facility’s power demands and capacity to cope with an increased workload.

In order to optimise the data centre successfully, IT leaders and engineers must address three key areas:

Analysis of power needs

• The first step engineers should take, if this is not already in place, is to conduct a review of power utilisation across the data centre/s as well as power demand from different business units. This initial step could reveal unused power capacity.
• Business units tend to ‘hedge’ their power requests — asking for around 30% more than they actually require. Therefore, the engineering team should have the appropriate internal approach to ‘test’ business units’ requests for power to calculate how realistic they are.
• For instance, a review process could use ‘volumetric analysis’ to effectively capture data which reveals how much power is really needed by different business units.
• This process calls for effective collaboration between the IT team, engineers and business units, as well as accurate data capture of power needs.

Implementing effective expansion

• Companies should be looking to expand by areas based on kilowatts of IT load — not by area required in square feet.
• Organisations should build outwards in an optimised incremental fashion — it is extremely inefficient to build a huge facility that will take years to reach full running capacity.

Ongoing energy monitoring

• Engineers should look to obtain accurate, ongoing insight into power usage throughout the data centre. For example, meter power usage at multiple points as far as the power distribution units (PDUs) and deliver this information to the customer for further analysis and reporting.
• By measuring to such detail, engineers can effectively optimise and continually improve power usage effectiveness (PUE) at existing and new facilities.

Expanding on the point regarding effective expansion, it is critical to look at the projected needs of the data centre in an incremental fashion, rather than building a huge box and waiting for it to fill.

Yes, data centre facilities are buildings, and buildings are real estate, and square footage is the universal language of real estate. But the lifeblood of a data centre is power and the conversation needs to start and end with that or a data centre project can go astray.

This demand is a direct result of the increasing strategic importance of data centres, led by a rise in virtualisation and a continual growth in demand for data storage.

In today’s high-density computing environments, power is the real determinate of data centre requirements and the best mode of expressing them in terms of kW of IT load, or, in other words, the kW of data centre-delivered electrical power required to drive the facility’s computing devices (mainly servers).

As a measure of power and capacity, kW most directly translates into the environment required to support the volume of MIPS and terabytes that the data centre is to support both now, and in the future. What’s more, it is a real, finite measure — there is no functional obsolescence of kW of IT load.

Cloudera’s Sean Anderson says of the conundrum: “Clear alignment with business goals is the key factor in moving forward. Understanding the technology is one aspect to making an integral shift in your information architecture, but it’s not the only one. A company can partner with a vendor, engage a systems integrator or hire the right people to ensure they are successful with the technology.

“A much more complex challenge is changing the culture of an organisation to not only be successful with new technology but also ensure that you have the right minds thinking the same way about how data is supposed to fuel the business,” he said.

As each successive generation of computing and storage gets more efficient per kW, the data centre gets more MIPS/terabytes out of the same level of kW of IT load. Therefore, any evaluation on the viability of expanding a data centre must come back to the estimated power load that will be placed on the increased space, and the availability (and reliability) of that power.

As data centres continue to expand in Australia to meet this rising demand for more computing power, engineers, electricians and cablers will work closely with IT leaders to compute the optimal balance between size, power and computing performance.

Image credit: © iStockphoto.com/Baran Özdemir