Interesting debate at the Data Centre Efficiency event that the SusteTECH project has put on in Bristol. John Milner from the JISC laid out the case for the JISC run, HEFCE inspired Universities Modernisation Fund in providing a brokerage service that will deliver cost efficient cloud resources for the sector. The debate is around whether this type of provision can match a properly designed and engineered local institutional data centre.
http://www.jisc.ac.uk/whatwedo/programmes/umf.aspx
Another pint being made is that PUE is a bit of a trap as a measurement of data-centre efficiency. Because PUE is a simple ratio of IT energy load against total energy load for the data centre, it can hide the fact that one way to tackle the question of data centre energy use is to start to address the IT load. How can we do this? Projects like the Cardiff University Planet Filestore project show one approach to reducing the load by shifting seldom accessed files to lower tiered storage, but there are other approaches that might provide some ways to tackle this. One is to look at the efficiency of the software and its use of computing resources, another to look at the overall provision of IT – and here the work of JISC’s Flexible Service Delivery Programme has relevance as it is helping institutions to makes sure that their IT provision is actually aligned to the needs of the business.
All in all, a very stimulating and enjoyable morning, with a visit to the University of Bristol’s HPC facility to round things off.
http://www.jisc.ac.uk/whatwedo/programmes/umf.aspx
Another pint being made is that PUE is a bit of a trap as a measurement of data-centre efficiency. Because PUE is a simple ratio of IT energy load against total energy load for the data centre, it can hide the fact that one way to tackle the question of data centre energy use is to start to address the IT load. How can we do this? Projects like the Cardiff University Planet Filestore project show one approach to reducing the load by shifting seldom accessed files to lower tiered storage, but there are other approaches that might provide some ways to tackle this. One is to look at the efficiency of the software and its use of computing resources, another to look at the overall provision of IT – and here the work of JISC’s Flexible Service Delivery Programme has relevance as it is helping institutions to makes sure that their IT provision is actually aligned to the needs of the business.
All in all, a very stimulating and enjoyable morning, with a visit to the University of Bristol’s HPC facility to round things off.
Professor John Seddon, an expert in service organizations with extensive experience in public sector systems says that there are two arguments for sharing services. The ‘less of a common resource' argument and the ‘efficiency through industrialisation' argument.
ReplyDeleteThe former argument is ‘obvious': if you have fewer managers, IT systems, buildings etc; if you use less of some resource, it will reduce costs. But the reductions are often minor and one-off.
The second argument is ‘efficiency through industrialisation’. This argument assumes that efficiencies follow from specialisation and standardisation – resulting in the creation of ‘front' and ‘back' offices. The typical method is to simplify, standardise and then centralise, using an IT ‘solution' as the means.
The problem with the industrial design is simple - it doesn't absorb variety in demand. Because of this, costs soar as the IT system has to be modified and customers ring back again and again because they can't get what they want.
The evidence of this flawed theory can be found everywhere. In HMRC or South West One shared services which predicted savings of £176 million over 7 years and actually recorded a pre-tax loss over its three financial years. Duplicate payments sitting at £772,000 and a struggle to manage £12.9m in outstanding debts.
This week Western Australia followed Queensland in ending its shared services. It was claimed that it would save $58 million a year and instead cost $444 million dollars (no savings). It is estimated that it will cost taxpayers between $1 - $2 billion dollars to rectify.
Howard Clark
The Systems Thinking Review