Data warehousing: upgrade, extend or replace?
The B-eye Network recently conducted a survey on behalf of Dataupia. which asked, among other things, companies whether they would consider a "rip and replace" approach to solving whatever data warehousing problems they might have. Seventy-five per cent of respondents said no.
Now, apart from the fact that I find the term "rip and replace" pejorative this raises some interesting points. The first is that I find 75 per cent surprisingly low. It suggests that fully a quarter of all organisations are so fed up with the problems and costs associated with their current solutions (if "solution" is the right word for something that is obviously failing to deliver) that they are prepared to go through all the upheaval of ripping out and replacing their current systems.
However, it is not on this aspect that I want to focus but rather on the 75 per cent who are loyal, or potentially loyal, to their existing data warehouse supplier. Now, we must suppose that there are some such customers that are happy with their current lot: they haven't got increasing data volumes, are happy with the amount of data that is currently archived to tape that can't be queried, they don't have any demand to embed business intelligence capabilities into operational applications, the performance of their queries is so fast that they get returned within the blink of an eye, they have no concerns over the ease and cost of administration of the existing system, there are no additional users that want to be able to use the warehouse and there is no requirement for supporting unpredictable or complex queries that might overturn the performance applecart.
Well, leaving them aside, what can any normal company, which has exactly these sorts of issues, do? If replacing the existing system with something bigger (or smaller in some instances) and better is not an option?
The first thing is that depends exactly what your issues are. If your issue is primarily one of performance then the obvious option is to add one or more data marts: presumably from a data warehouse appliance vendor, which is fine though it may mean that your whole environment becomes more complex.
However, if your issues are more complex: for example, you have concurrency and data volume issues as well as performance, then simply adding some new data warehouse appliances is unlikely to resolve the problem and, given that you don't want to replace your existing system, then you are going to have to improve the existing system in some way.
There are two ways of doing this: either you can upgrade your existing system through a new version of your supplier's software or by upgrading the hardware that it runs on, or you can extend your existing environment via a third party.
The problem with upgrading the existing environment is that what you get is more of the same. Certainly, if you upgrade to Oracle 11g, for example, then you can use its compression features to reduce your storage requirement, and you will get some performance benefits too. In other words, you will get incremental benefits. In fact, you may even get significant benefits but what you probably won't get is the sort of order of magnitude benefits that you want.
And it is these companies, who want order of magnitude benefits but don't want to rip and replace, that Dataupia is targeting. Hence its sponsorship of this survey. So, can it deliver on that promise? Well, it is early days for the company but one of its first customers, which is using Dataupia in conjunction with an Oracle database is now loading 30-90 days worth of call data records (it is a telecommunications company) into its warehouse, whereas previously it only stored them for between three and seven days. And it is getting the same sort of performance on the expanded system as it did before. That sounds like an order of magnitude improvement to me.
Philip Howard, is Director of Research - Technology, at Bloor Research.
Copyright © 2007, IT-Analysis.com
Sponsored: Benefits from the lessons learned in HPC