We are looking at how best to store data so that we can map trends over months and years, in what is known as a data warehouse. We do have some digital records already, for example we can looks at trends in student numbers for over a decade. With more and more processes online, the University administration want to look at trends over a wider range of data, and to ask questions of data that go beyond the individual administration units.
This is not trivial, especially as we don't know which questions people will want to ask in the future. We need a flexible data model that allows reports to link data that was taken from more than one operational system. For example, a question about the cost of running any particular course would need to know about the cost of the rooms used, the staff, library materials, any special equipment, and so on.
We have looked at a number of ways of tackling this problem. The approach we are now proposing is to have two levels in our data warehouse. The first will be a very flexible data model that can integrate information from all sorts of systems. The second will present greatly simplified views of the data that address a particular set of questions - such as student numbers, or course costs, or research income, and so forth.
The diagram shows the end-to-end model, as it would apply to users in the Estates department. They would see the operational systems that they use day-to-day (shown on the left), and the strategic reports and dashboards.relevant to their work (shown on the right). Inbetween, the data warehouse will integrate this data with information from other areas of the University, and preserve it over time.
In a more technical forum, I will write some notes about the details of the data modelling approach. It's the sort of thing that is very interesting to technical people and rather less so to the people who just want to use the system.
This is not trivial, especially as we don't know which questions people will want to ask in the future. We need a flexible data model that allows reports to link data that was taken from more than one operational system. For example, a question about the cost of running any particular course would need to know about the cost of the rooms used, the staff, library materials, any special equipment, and so on.
We have looked at a number of ways of tackling this problem. The approach we are now proposing is to have two levels in our data warehouse. The first will be a very flexible data model that can integrate information from all sorts of systems. The second will present greatly simplified views of the data that address a particular set of questions - such as student numbers, or course costs, or research income, and so forth.
The diagram shows the end-to-end model, as it would apply to users in the Estates department. They would see the operational systems that they use day-to-day (shown on the left), and the strategic reports and dashboards.relevant to their work (shown on the right). Inbetween, the data warehouse will integrate this data with information from other areas of the University, and preserve it over time.
In a more technical forum, I will write some notes about the details of the data modelling approach. It's the sort of thing that is very interesting to technical people and rather less so to the people who just want to use the system.
Comments