The evolution of the datacentre

Mark Ridley, director of technology at recruitment web site Reed.co.uk, is exploring the possibility of moving datacentre systems to the cloud to improve efficiency, reduce costs and increase flexibility.

“Reed has experienced an evolution of the datacentre rather than a revolution. We have come a long way since 2000 when we had a server room rather than a datacentre, to the point where we are now using virtualisation technologies and contemplating cloud computing for our 2010 upgrade,” says Ridley.

Reed’s original datacentre was rudimentary and “not an outcome based on a lot of thinking”, admits Ridley, but his appreciation of the power of the web and what it could do for the business led Ridley to focus on web application development, while a third party was brought on board to run the datacentre.

“Development, some of which was outsourced, was brought entirely in-house as we realised we had to invest more in online job posting, while Attenda became our hosting partner. Our core strength is development and understanding why and how the business wants to do things, but we have outsourced the datacentre to experts so we can access technology we couldn’t otherwise afford,” he says.

Ridley regards Attenda as a partner rather than a supplier, but investigates alternative providers as part of Reed’s hosting re-tendering process, which it undertakes every three years.

“Our last refresh was in 2007, but two years into the contract, we began looking. In the past we found we quickly hit our headroom for the server we had implemented, but this time we are not in that position thanks to the flexibility of virtualisation technology, which was introduced in 2007,” says Ridley.

Investment in standardisation of processes, procedures and platform has helped lower datacentre costs for Reed and improved performance. “Guidelines and policies are followed so there is a great deal of consistency of service, and Attenda knows how to deal with any incident,” says Ridley.

Attenda architects work with Reed to identify new areas where virtualisation can be employed.

Currently, Ridley is working with the supplier to implement a new VMware-based virtual server to run its StrongMail email system.

“We send about 25 million emails a month and StrongMail has improved our email delivery rate from 90 to 99 per cent. This has improved the efficiency of our datacentre as sending 2.25 million emails that don’t reach their intended recipient wastes processing power. We are always looking at ways to improve processes and energy efficiency and using a virtual machine to host the StrongMail box would make further energy savings,” says Ridley.

A green and efficient datacentre is important to Ridley, who is always looking for ways to cut his power bills.

“By ensuring good performance, you need fewer servers and less power. For example, a badly coded application will require high bandwidth and many servers,” he says.

In the past, Ridley has brought in-house any servers that are freed up at the end of its three-year contract with Attenda to run the company’s development environment, but now he benefits from using virtualisation.

“We bought three new HP DL380 servers and created a brand new, swanky virtualised environment that uses half the energy required by our old inherited servers,” says Ridley.

Over the years, the firm’s datacentre strategy has evolved. "In the early days, we wanted to own all the kit. Pre-2007, we owned the storage and firewalls, but now we make use of the common infrastructure shared across Attenda clients. This gives us access to superior kit, for example the storage area network (SAN) that uses virtualisation technology from VMware. We buy chunks of storage, which is cheaper and better performing than if we had to invest in dedicated storage,” says Ridley.

Non-ownership reduces management costs and increases access to external skills so Reed can concentrate on its technology priorities, he says.

“If a disk drive fails at 3am, we don’t need to know about it unless there has been a degradation of service affecting our service level agreements. A good hosting company has someone who lives and breathes storage and there is no way we would have that expertise on hand at Reed. We need to focus on development,” says Ridley.

Reed’s database cluster and web servers, however, have dedicated hardware and are run within a ring-fenced area in the Attenda hosting environment, but Ridley can foresee this changing based on his positive experience of virtualisation.

“Because our web servers and database are so critical to what Reed is delivering, we did not want to opt for a shared infrastructure. However, virtualisation and shared applications will change the way we buy datacentre power. I can envisage a cloud computing model where we say how much power we require without worrying about hardware. In the past I have had to trundle with a server in a white van across London, but that need not happen now,” says Ridley, who is looking at cloud computing for the 2010 datacentre upgrade.

“In the future, we will not be looking at the provision of IT as physical boxes, but the provision of a service to a client as a unit of computing power, and the expectation will be that it comes on just as water comes out of a tap when it is turned on,” he adds.

Cloud computing will especially benefit businesses such as Reed that are subject to spikes of demand, Ridley believes.

“Our busy time is the New Year, when many people look for new jobs. In the past, a big burst of activity has meant our developers had to throttle the number of people coming through to the site. With cloud computing we would be able to simply phone up and order more computing power for January,” says Ridley.

Cloud computing could also lead to shorter datacentre contracts. “We will not be tied into hardware costs –­ the hardware will belong to our hosting provider,” he says.

Another advantage, Ridley foresees, is not having to make guesses about the economy three years in advance and how it will affect use when ordering computing power.

“With a cloud computing model within the datacentre, IT provision will become more timely, efficient and easier to budget for as it will be aligned with what the technology group does with the business,” says Ridley.

Read our case studies to find out how Bolton Wanderers FC and Nottinghamshire Healthcare Trust manage their datacentres

Six steps to greater datacentre efficiency
To help IT professionals improve datacentre management while cutting costs and improving reliability and resiliency, Forrester Research's Doug Washburn offers the following tips:

Rationalise your application portfolio. The datacentre and its expensive and energy-consuming IT, power distribution and cooling equipment exists to support applications. And it’s not uncommon for applications to be severely under-used, which snowballs into IT infrastructure capital and operating expenses. Datacentre managers should liaise with users on the phasing out of severely under-utilised applications.

Consolidate and eliminate under-utilised servers. Datacentres are plagued with “dead” servers, or those with utilisation levels below six per cent, consuming power, cooling and space resources. Eliminating and consolidating these servers will free up capital and operating costs, and extend datacentre life by freeing up power, cooling and space capacity.

Increase your server virtualisation ratio. While nearly every organisation admits to virtualising servers, the savings potential from virtualisation is often not fully realised. Forrester finds that mature virtual production environments usually have about 30 virtual machines – and coupled with advanced automation tools, this could reach 50, without undermining service-level agreements.

Enforce “virtual first” policies for new applications. By stipulating that all new applications must run on virtualised infrastructure, you will benefit from improved disaster recovery and business continuity; rapid — or even automatic — restart of applications after an IT failure; and when used in conjunction with data replication between datacentres, it can restart applications at a recovery site following a primary site failure.

Increase storage utilisation and reclaim storage capacity. Storage environments are plagued with low utilisation rates and highly redundant data. IT professionals should consider thin provisioning and data deduplication technologies to improve utilisation and reclaim storage capacity.

Optimise your datacentre temperature. While manufacturers of IT equipment have set the allowable high-end temperature at 27°C, most datacentres are too cold, operating at 8°C to 20°C. With 60-70 per cent of datacentre energy consumption going to power and cooling, this represents a significant operating cost. Under supervision, turn up the temperature in your datacentre. For example, one IT manager took his datacentre temperature from 20°C to 23°C and recorded a 12.7 per cent reduction in energy use.

Doug Washburn is an infrastructure and operations analyst at Forrester Research. Several Forrester reports are available free of charge to Computing readers by visiting www.forrester.com/computinguk.