API WarsWeb services platforms — .Net and J2EE — take the fight to the analytic arenaby Seth Grimes Continued from Page 1 No wonder then that data mining is slated for coverage in XMLA: Microsoft needs a .Net migration path for OLE DB for Data Mining (ODDM), which was folded into Microsoft Analysis Services and would otherwise be orphaned. The November XMLA Council meeting saw the start of efforts to draw an XMLA-data mining roadmap. Perhaps that roadmap will lead to inclusion of data mining techniques in the analytic toolkit for everyday use; data mining has yet to escape from niche applications. Technically focused efforts to date have failed to help ODDM break out. Notably, leading data mining vendors have shunned ODDM, and while they've adopted Predictive Modeling Markup Language (PMML), they've had to "heavy it up" with proprietary extensions to make it work. The promise is supposed to be model interchange between applications, for instance allowing an analytic engine to score a credit application using a model that a data-mining tool created and exported. The reality, according to Ted Morris, an SPSS Inc. engineer, is that differences in algorithm implementations force users to add control parameters and sacrifice interoperability. Ted told me that he's "coming to understand that [PMML] is so limited in scope that it's an academic exercise." Oracle's Mark Hornick is convinced that PMML extensions will be standardized and that the advantages of enabling interchange between model-building and scoring applications are real and important. And the reality is that vendors including IBM, Oracle, SAS, and SPSS have successfully demonstrated PMML-mediated model interchange. There's additional good news on this score: PMML should be supported within both JDM and XMLA. Motives vs. ResultsAlthough I've adopted a skeptical tone in assessing the motives of various analytic computing vendors, I don't wish to imply that the machinations that have produced the various APIs have tainted the outcome. On the contrary, strong motives and a deliberate, if self-interested, approach can ensure well-defined, usable results. I've given my take on Microsoft's motives and directions and also on those of Oracle, the anti-Microsoft. Certain other companies are closely allied with one camp or another, such as Microsoft Analysis Services client and toolkit providers Crystal Decisions, Panorama Software Systems Ltd., ProClarity Corp., and Simba Technologies Inc. Their solutions are very well suited to certain markets. I've cited other companies that sit in both camps: Hyperion and SAS — larger companies that seek to maximize interoperability to maintain and expand their market standings and position themselves to dominate the emerging business performance management (BPM) space. Effective BPM will provide cross-functional analytics across disparate business systems; integration is key. A few noteworthy vendors including Alphablox Corp. and Informatica are sitting out the fray. Representatives of both companies use similar language. Alphablox product and marketing vice president Bill Wagstaff told me that he and his company "take a very pragmatic approach. [They] recognize the need to interface ... but have a very rich abstraction layer and can afford to wait for the market to mature." And Ivan Chong, senior director of product marketing at Informatica says that his company is "very eager to adopt the [emerging] standards" but has "a more utilitarian approach based on customer need." Chong said that right now, customers view standards battles "as vendors jockeying for position." I see smaller, independent software vendors and nonvendor developers facing the toughest questions: Do you go with XMLA, relying on Microsoft's market strength but limiting yourself to .Net analytics providers and the MDX language and its successors? Can you weather Yukon and .Net-platform rollout delays? By contrast, does the richness of the J2EE platform and the metadata and communications standards that complement core OLAP and data-mining APIs merit the commitment to Java? The battlefield here is analytics; similar questions apply across the larger Web services world. Most of us can simply sit back and watch the .Net vs. Java fight unroll. It's likely that both platforms will thrive and that Microsoft won't win the platform hegemony it has sought for many years. Thus, we'll see continued competition, which will promote ongoing innovation even if it disappoints those who hope for one universal analytic API. Seth Grimes [grimes@altaplana.com] is a principal of Alta Plana Corp., a Washington, D.C.-based consultancy specializing in analytic computing, business intelligence, and demographic, social, and economic statistics. RESOURCESJava Data Mining (JDM) specification: jcp.org/en/jsr/detail?id=73 Java Metadata Interface (JMI) specification: jcp.org/en/jsr/detail?id=40 Java OLAP (JOLAP) specification: jcp.org/en/jsr/detail?id=69 Object Management Group, including Common Warehouse Metamodel: www.omg.org Predictive Modeling Markup Language: www.dmg.org XML for Analysis Council, including XMLA specification: xmla.org Related Article at IntelligentEnterprise.com: "XML for Analysis Decoded," Aug. 31, 2001
|
 |
Most Popular This Week
IE Weekly Newsletter
Subscribe to the newsletter
| |
| ||||||||||||||||||||||||||||||||
