Your bumpy path toward getting a clear, single view of the customer by using a data warehouse can be smoothed with the right kind of database.
Part of the complexity begins with efforts to coalesce disparate data that resides in databases linked to already functioning applications. In addition, much of the valuable data inside an enterprise is unstructured and not well conditioned for traditional relational databases.
Vendors also underestimate the challenge of supporting a data warehouse as compared with more typical transactional processing.
NCR Corp. Teradata focuses its database resources specifically for data warehouse workloads.
Teradata’s database uses a hash-based file system in contrast to the index-based system favoured by companies such as IBM Corp. and Oracle Corp. and used mainly for finding rows in a transaction process.
An index-based system works well for write-oriented tasks requiring low latency for juggling lots of transactions concurrently; Teradata’s product isn’t designed for these kinds of tasks.
For read-oriented analytical operations requiring a multitude of data connections, Teradata’s database is eye-catching. Wal-Mart, Bank of America, 3M and British Airways have been enticed into using Teradata’s technology to consolidate data marts into a data warehouse.
As for competitors, says Kevin Strange, an analyst at Gartner, there’s “little evidence” Oracle is improving its ability to handle complex application-neutral data model implementations that are at the heart of data warehouse activities.
Strange says there’s some confusion over IBM’s approach: Should companies choose the extended DB2 product or Red Brick and the Informix Extended Parallel Server (XPS)? “IBM sales teams have been proposing Informix XPS and Red Brick in some new customer implementations, even with IBM management stating these products would not be proposed for new implementations,” says Strange.
Microsoft’s SQL Server has received low-end acceptance for simple data marts but lags when faced with large data volumes and complex data models.
Sybase’s IQ and Adaptive Server Enterprise products are candidates for database marketing where the organizing of customer lists is based on only a few criteria.
But a data warehouse is something else. It must be powerful and versatile enough to give non-IT workers the chance to analyse data without IT constraints. There’s little value in analysing data that has in essence been predigested because the database can’t handle complex queries.
But remember, if you can’t apply parallel-processing techniques to your workload, hardware boosts will be a waste of money. And that would make your road to a single customer view a bumpy one, indeed.
Pimm Fox is Computerworld (US)’s West Coast bureau chief. Contact him atpimm_fox@computerworld.com.
