SQL Insider: Data Mining Goes Mobile

Is there a market for a SQL Server lite?

I'll concede that most of you aren't interested in doing hard-core data mining—the kind involving six-figure budgets and a small army of specialists with PhDs in statistics or applied mathematics, or expertise in neural networks. However, I suspect many of you have a market for a true SQL Server lite that could run across multiple Personal Digital Assistant (PDA) platforms, such as PalmPilot and Windows CE (WinCE) devices, and that could be used in embedded systems. Sybase, with its SQL Adaptive Server Anywhere (ASA) and new UltraLite deployment technology, took an early lead in the mobile market and continues to dominate the landscape. Oracle has been shipping Oracle Lite and has recently expanded its own lite line to support the Palm OS as well. Even IBM has a pair of new DB2 lite products called IBM DB2 Universal Database Satellite Edition and IBM DB2 Everywhere for WinCE and Palm OS, which are scheduled to ship in the third and fourth quarters of this year.

However, as far as data mining goes, Microsoft seems to be borrowing a page from its own playbook—leading an industry effort to develop extensions to OLE DB, just as it did with the OLE DB for OLAP multidimensional extensions (MDX). Of course, the interesting question is whether Microsoft will ship a basic data- mining services engine à la OLAP Services. I'd certainly like to see a basic mining engine for text and data—leaving specialized mining of audio and image data, for example, to the independent software vendors (ISVs). But imagine a basic engine that could, in the text domain, automatically generate document summaries and classify documents by content. Today, you'll spend at least five figures to achieve that functionality from IBM's Intelligent Miner for Text or Oracle's ConText.

Also, Microsoft seems surprisingly silent about geographic information systems (GIS). Perhaps Microsoft was so focused on demonstrating SQL Server's scalability to the realms of very large databases (VLDBs) when it developed and deployed its Terraserver site that it neglected to pitch SQL Server as a database management system (DBMS) for GIS applications. You don't hear much about how people are using SQL Server to develop GIS applications. Yet, according to Environmental Systems Research Institute (ESRI), several thousand sites are running SQL Server with ESRI's Spatial Database Engine (SDE). Several tens of thousands of sites are running Oracle or DB2 databases with SDE, but my point is that SQL Server is ready for GIS prime time. Also, ESRI and MapInfo, another leading GIS vendor, have OLE controls that make it easy for Visual Basic (VB) programmers to create map-enabled applications.

Benchmark Boom

Previous issues have covered the ongoing database benchmark wars, featuring Oracle, Microsoft, and IBM. To date, most database benchmarks are the Transaction Processing Council's TPC-B and TPC-C benchmarks, developed to predict online transaction processing (OLTP) database performance. The relatively new TPC-D benchmark is better suited for decision-support and business intelligence (BI) applications than the earlier TPC benchmarks were, and the OLAP Council also has developed a new BI-oriented benchmark called Analytical Processing Benchmark (APB-1), which is now in Release II. For more information about the politics behind OLAP benchmarks and APIs, read Nigel Pendse's commentary at http://www.olapreport.com. Also see Michael Otey's editorials on benchmarks in the April and June issues of SQL Server Magazine.

Hide comments


  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.