Costing IT services

This is a question that many are asking and for which, as yet, there are no clear answers or even proven ways of finding out. As part of the team at Jisc looking at this problem, we have some approaches and are developing tools that can help you both benchmark your costs against others and drill down into the real costs.

Energy costs

There has not always been a level playing field when it comes to IT costs, perhaps most clearly in relation to energy costs. 

Our studies have shown that investment decisions, that might reduce IT energy use and associated carbon emissions, are being hampered by the lack of clarity about the energy being used by institutional IT and the lack of mechanisms to reward these investments with a share of the costs saved. In short, the IT department would typically spend the money on energy efficient equipment, while the facilities department would reap the benefit through lower utility bills.

Our recent report explores the issues in some depth and shares examples of how institutions have addressed them.

Operating costs

This lack of transparency in IT costs extends further than just the cost of energy.

The EU funded e-Infranet project has highlighted, in a report to the EU Commission, that energy and associated infrastructure costs of data centres (cooling, power supply etc) can be as much as 50% of operating expenditure and a big proportion of capital costs. However, these costs are often hidden from view as they may be counted under wider categories or lie outside the IT budget altogether. This can make decisions about shared service solutions or moving to cloud based services difficult to make, as these other options will typically have all costs accounted for. This can make alternative models of IT provision look expensive.

Meeting of minds

Last month (June 2013) as part of our work in this area, university IT directors, data centre and high performance computing managers, commercial data centre operators and finance experts met to debate these issues at the University of Liverpool.

The details of which can be seen in this rich media mash-up.  Some of the points that struck me as relevant and I think important to the sector are:

• There are no right answers in choosing between the various cloud and hosted offerings for IT services – decisions between the trade-offs between costs, risk and appropriate technological fit are as hard to balance as they have always been. However, understanding the real costs of your IT services is a good place to start. The University of the Creative Arts found a good balance between different suppliers and models of delivery in their procurement of a mixed hosted/cloud data centre.
• Tracking the real costs of IT services in a university is hard, as the University of York explained  – you need a firm grasp of the services you deliver at a detailed granular level, which takes time and effort to assemble.  Without time sheet data, allocating staff costs is pretty much a stab in the dark, at best, and allocating truly infrastructural costs like network provision is difficult. For example, how do you go about determining the network costs of the student record system?
• Benchmarking against your peers is a useful exercise and our Financial X-Ray tool is one way to do this.
• It is possible to construct incentive schemes to bring awareness of energy costs to data centre end users. At the Department of Engineering of the University of Cambridge, a re-engineered data centre makes use of free air and evaporative cooling approaches to achieve an impressively low Power Usage Effectiveness (PUE) measure of 1.1. Part of the stimulus for this development came from the university’s financial incentive scheme. Departments are given an annual (reducing) allowance of energy and get a small cash bonus if they undershoot, but pay a penalty if they overshoot. At the Department of Engineering they do not recharge the users for the energy they use in the data centre as yet, but they report that the very fact of the existence of the incentive scheme and the way that the department makes available its energy use performance figures has made the users take active steps to reduce energy use.
• True efficiency in data centres requires optimisation in multiple dimensions. Following best practice from the EU code of conduct for data centres or the Green Grid or our own innovation work with data centres can help tackle the facility costs but that will not tackle the inefficiencies inherent in the computers themselves.
• Developments like direct liquid cooling will make a real difference in the future, as will improvements in chip design and optimisation of code to run more efficiently. We must move from thinking in terms of a crude power usage effectiveness (PUE) measure to think about ‘watts per FLOP (Floating Point Operations Per Second)’ or even more precise metrics of how much useful work the computers in the data centre are actually doing. Higher education is in strong position to help with this, as they have data centres and engineers as well as computer scientists who can contribute to the development of this agenda.

One thing that really pleased me is that one of the participants, who presented on the energy incentive scheme at Cambridge, told me that he had attended a Jisc workshop on data centres five years ago at Cardiff, and it was partly as a result of what he had learned there that he had the courage to persist with the developments he was responsible for at Cambridge. His data centre is now an exemplar for the sector and he has many requests for visits and briefings from people keen to know how he did it.

Sometimes we cast our bread upon the waters…