We are at the University of York for a workshop to start the process of engagement around the developing outputs from the Costs of IT Services project (CITS).
The fundamental question is “how much does your IT cost?” Not, how much is in the budget, but how much does it actually cost. How much does this or that service really cost? The reason for asking these questions, apart from the obvious need to be on top of expenditure and in control of your destiny, is that it gets hard to make sensible judgements around central versus local provision, or in-house versus cloud solutions, if you do not know how much things are costing you in the first place!
Many questions got raised – like apportioning costs of things like the network. How do you allocate the network costs to, say, the student record system or email? Or should you bother? Is the network just something that you must provide because without it, nothing else would be possible?
The audience was a good mix of IT managers (some senior), HPC and Systems managers, business analyst, and various people occupying variations on finance-related roles. The meeting was hosted by Heidi Fraser-Krauss, IT Director at the University of York.
The University of York have done some of the hard thinking on this subject, as primary fieldwork partners for the Costing IT Services project. One of the issues they faces is where to stop counting. York has quite a small central IT team, so much development of existing systems actually happens in the departments, while IT Services run the infrastructure that makes these things work. For example, IT services run the computers that power the student record system, but development work on that system is the responsibility of the Registry. So what do you count to get a true picture of the cost of that service? And cost to whom? Just the cost to IT Services, or the cost to the university?
HPC and Research Computing
What are the models for providing support for research computing and and how does the cloud fit into this? How do we really compare the costs of cloud (which will be a commoditised price that includes the capital costs, running costs, cooling and power conditioning overhead, security, resilience and network costs, with in-house version where the costs of many of these items are either not available or are very hard to get hold of and will be shared with a myriad of other services?
What price would you put on being sure that your cloud provider will still be there next week? Peter James raised the example of Nirvanix, a US cloud provider that gave it’s users 14 days to remove their data as it went into bankruptcy this year.
One issue that surfaced in a number of points through the day was the danger of charging for IT services internally, as this is simply to make more likely that departments will simply go and get their own cheaper (at the point of purchase) IT kit, but then shout for help when things go bad on them.