A new trend has entered the Data Warehouse market place: the data warehouse appliance. The idea is to offer a complete solution for the data warehouse in a single purpose, dedicated box. It is completely integrated, pre-configured and pre-tested. Plug it in and you're ready to start loading data. At least: if you consider the hardware+software type of appliance. There are also software-only appliance, that are often packaged as virtual machines (VMWare or other). In that case you need to provide the iron yourself.
So it's analogy lies with houshold appliances. Please don't think: washing machine (as I've seen people do when talking about appliances).
Think: deep-freezer. Why? Well, we all should be aware that a data warehouse doesn't clean your data. It comes out just as dirty as you put it in. Data should be cleansed in the source systems, not in the data warehouse. That said, the data warehouse resembles a deep-freezer much better: anything you put in will remain there, intact. And if you organize it well, it will be fairly easy to find stuff later for retrieval. Of course, your 'frozen data' will be quite raw, not ready for consumption. So you'll need to process it before you can put it on the table (data mart analogy?) and serve it.
The benefits?
- simple to operate and maintain (at least: in theory)
- rapid time to value
- low cost
- high performance
To be honest, this is not an entirely new concept. For example TeraData has been around for many years now and can arguably be placed into this category.
What is new, is that many more vendors have recently entered the market place, boosted by advances in database technology (e.g. column-based databases) and vritualization of hardware (VMWare, Microsoft HyperV, Cloud computing).
All major software vendors (Oracle, HP, IBM, Microsoft) now have an applicance offering (most through acquisitions), except perhaps SAP. This will without doubt lead to a consolidation in this emerging market, but at the same time boost acceptance of the proposition.
Data warehouse appliances are definitely interesting, as they can take away some of the hassle associated with building a stack of compatible products (OS, ETL, database, etc) and have them configured to co-exist peacefully and efficiently. Still be aware that 90% of the cost related with data warehousing is related to analyzing and fixing the data (in the source systems!), determining the reporting and analysis needs and transforming the incoming data into useful information. That part of the work is not going away.