In the latest of Information Age’s roundtable debates, members deliberated one of IT’s enduring problems: how best to control the infrastructure.
The lunch debates are a regular forum for up to 20 specially selected CIOs to discuss the issues that dominate their corporate agenda, gaining insight into how some of the leading businesses are getting to grips with those issues. The lunches are held under the Chatham House Rule, which allows delegates to speak openly, without concern that their comments will be attributed to them or their company.
At the September event in London’s famous Rules restaurant, sponsored by APC, discussion centred on optimising the IT infrastructure. All delegates reported that they were, or had been, involved in data centre consolidation. The impetus came from a number of sources: need for control; price pressures; the shift to server-centric computing. Whatever the driver, all agreed that bringing the data centre under control, was a vital aspect of sorting out the infrastructure, and that it will remain a corporate priority.
The proliferation of systems throughout the 1990’s and into the early part of this decade has left many firms with a morass of IT infrastructure. Add to this the effect of multiple mergers and acquisitions and systems soon start to look complicated, in need of rationalising. One delegate reported company growth had saddled him with ten separate data centres.
Cost savings can also be a very powerful motivator for firms contemplating making their IT infrastructure simpler. For some it can appear the only option. One member of our panel described how he has been compelled to reduce his IT spend by 25% by 2008. In these circumstances, the target was not a joint agree-ment between IT and management board, but a company-wide cost saving target, leaving him with the challenge of meeting it. Optimising data centre spend looks like the best place to find the savings.
Similarly the CTO at a leading UK retailer estimated that he saved by 20% by standardising his server farms. “We’re trying to move to a utility pricing model, where we only pay for what we use,” he said.
Staff costs can also be an issue. One of the delegates, the CIO at a nationwide leisure group, said that consolidating its infrastructure into one centre manpower slashed costs. “Otherwise we have 45 [centres], each with its own systems administrator.”
However, cost savings are not the only driver of consolidation. Control through centralisation is also persuasive. Several of our panel had seen their companies go through a series of acquisitions. Through consolidating the number of data centres each organisation was able to impose standards across the business, reducing the time and cost required to support heterogeneous environments. “We’d expanded rapidly through acquisitions, but you need to know how to manage the new areas,” said one.
Despite considerable pressures to consolidate, our panel of eminent CIOs agree: There is no ‘one size fits all’ model for achieving this in data centres.
Firstly, you have to consider how far the organis-ation can consolidate. A single centre is also a single point of failure, and IT legacy may make it impossible to consolidate that far, said the CIO of a large industrial manufacturer.
Picking up on that theme, one delegate explained that the sheer number of records his company generated – between 18 and 20 million a-day – made the process of consolidation immensely tricky. The challenges of running parallel systems while under-taking the consolidation programme were too great a barrier to ever start out on the project, he said.
Outsourcing the infrastructure can appear an attractive option where you no longer need to worry about whether to run Wintel, Linux or proprietary boxes: everything comes down a browser, said the IT manager of a multinational chemicals manufacturer. “But you have to budget for 10% at your end to manage the service level agreements,” he said.
Consolidation can also generate other management headaches, including who is running your data centre, and what skills they possess. Several of our panel reported that there was an assumption that so-called ‘commodity’ servers only needed low-level skills; in fact the job of managing these server farms requires mainframe type skills.