In Praise of Bundled ETL

What are the economics of bundled ETL, and just what do Oracle and Microsoft bring to the table?

For last month’s Oracle Warehouse Builder (OWB) R2 announcement festivities, Oracle Corp. representatives singled out OWB’s new pro bono price tag as one of that product’s most significant selling points. Microsoft Corp. officials, too, have made much of SQL Server Integration Service’s (SSIS) shockingly low price point—it’s free for use with SQL Server, they argue.

Both claims are true—with an important caveat. Assuming that you’ve already ponied up the cash for Oracle 10g, SQL Server 2000, or SQL Server 2005, you’ve got carte blanche to use their ETL tools, too. But that’s a significant caveat, for a number of reasons. For one thing, neither Oracle 10g nor SQL Server is free. You’ve first got to purchase, deploy, manage, and—most importantly—pay ongoing maintenance in support of both database platforms. To describe either OWB or SSIS as free on that basis is a little like saying the transmission that came with your car is “free”—more or less.

Secondly, both relational databases are optimized for use in their own, mostly homogeneous environments. This has important ramifications for users who want to tap them as bona-fide “enterprise” ETL tools. Microsoft says SQL Server 2005 ships with canned support for connectivity into Oracle, DB2 (residing on the mainframe or AS/400), and other platforms, but OWB’s out-of-the-box heterogeneous connectivity is limited to ODBC and JDBC. In this respect, while the core OWB R2 might be “free,” customers who plan to make serious use of it in conjunction with other, non-Oracle data sources will in many cases opt to purchase Oracle’s Enterprise ETL Option, as well as Oracle’s platform-specific adapters (Oracle Transaction Gateways) for Siebel, PeopleSoft, and SAP data.

The problem, says Philip Russom, senior manager of research and services with TDWI, is that ODBC and JDBC—which are essential connectivity mechanisms, insofar as they’re both almost universally supported—are far from ideal, at least in so precarious a business as the extraction, transformation, and loading of enterprise data.

“Handy as these common standards are, they take a lowest common denominator approach to data access. A native gateway is far better, because it taps into a platform’s non-standard but richer capabilities,” notes Russom. “To achieve this kind of performance and extended options with OWB, users must buy additional licenses for gateways to non-Oracle source and target platforms. These gateways are not free, so OWB cannot be free.”

Not that this is a completely unprecedented caveat, of course. In fact, Russom observes, Oracle’s position is actually consistent with that of its ETL industry competitors. “Most ETL tools are licensed as a base model, with many add-ons. Gateways and other interfaces are typical add-on products. So OWB is consistent with market-wide licensing practice,” Russom says.” Experienced users know to expect increasing costs—in most cases—as they tie more types of source and target systems into their ETL infrastructure.” There’s another important catch, however—starting first and foremost with OWB’s dependence on Oracle’s bread-and-butter PL/SQL. “OWB is an application that runs atop the Oracle Data Server. Most of its functions generate PL/SQL code, which can only run on an Oracle Data Server,” Russom points out.

That being said, Russom and other analysts say that both OWB and SQL Server (and especially the new SQL Server 2005, with its revamped SSIS) can almost certainly benefit Oracle and Microsoft customers—even if they don’t plan on tapping either product for use as an enterprise ETL tool.

In this respect, argues Mike Schiff, a principal with MAS Strategies and a member of TDWI’s extended research collaborative, it’s all about pricing pressure. “Given that you’ve already bought the database, you’ve got to ask yourself, why do I want a third-party product? What is the point of a third-party product? At the very minimum, if I were any end user out there, if it’s like Ascential or IBM or Informatica making their pitch to me, I’d ask them: ‘Why should I pay a quarter of a million dollars for your tool when I can get it basically free with Microsoft or Oracle?” Schiff argues. “I would use it even if I wouldn’t consider using it to put pricing pressure on whoever I’m negotiating with.”

This argument becomes slightly less persuasive when both parties (i.e., users and the ETL vendors with which they’re at loggerheads) know that a proposed integration scenario involves multiple, heterogeneous data sources, but Schiff says it’s only a matter of time before the ETL pure-plays start to feel the heat.

“How long are the independent ETL vendors going to be around? Ascential’s already been picked up by IBM; Informatica is basically out there alone. It’s one thing if it’s just Oracle to Oracle, but as soon as these [relational database] vendors start populating databases other than their own, if they do it inexpensively, that’s going to put incredible pressure [on the ETL pure-plays].”

To some extent, Microsoft’s SSIS already does that—for Microsoft shops, anyway. “They [the ETL pure-plays] have always said, ‘We’re a general purpose tool; Microsoft or Oracle works best with Microsoft or Oracle.’ But we’ve seen Microsoft already lets SQL Server [2005] connect to non-Microsoft data sources, and as these [relational database] vendors start opening up to populating databases other than their own, that puts pressure on the independents.”

About the Author

Stephen Swoyer is a Nashville, TN-based freelance journalist who writes about technology.