Every writing format has its limitations, including this somewhat stovepiped, stand-alone one. To partially compensate, I'll try something different this month: updates of several of my past columns in the area of database management.
Database management system services represent where I think enterprise data architectures are going over the next two to 20 years. (Exception: small companies that buy complete application systems from single vendors.) DBMS2's core principles are:
1. Enterprise databases will never be centralized and will never have simple or comprehensive designs. This has been true for decades, and will long continue to be, for manifold reasons. Among them: Enterprises have legacy applications and get new ones when they acquire companies. They have packaged apps from multiple vendors; software as a service may accelerate this trend. New kinds of analysis are being invented all the time, requiring new kinds of data to be tracked that weren't envisioned in the original schemas and that may be more opinion than "facts" anyway. And really sealing the case against centralized DBMS is the fact that in a wide range of cases, focused data management products greatly outperform general-purpose ones in speed, cost or effectiveness.
2. Applications will be built on top of simple Web services. Today's XML-based Web services may be superseded by something similar yet sleeker. But the general idea is a winner -- databases and applications alike will be increasingly modularized and accessed via simple data-oriented interfaces. And while I'm not sure that the current buzzphrase "composite applications" will or should survive, I wholeheartedly endorse its general idea.
3. Eventually, these Web services will be managed through appropriate repositories. This part of the DBMS2 story will come fully true only in the long term. "Active repositories" have produced a long list of failures, from IBM's AD/Cycle to Oracle's Sedona and beyond. Really strong Web services directories shouldn't be expected until sometime in the next decade.
As soon as I introduced it in my August 2005 column "Time for a New View of Database Management," the DBMS2 idea met scathing criticism from a handful of relational database purists. But those who do the real IT heavy lifting have been much friendlier to it. Although using different terminology, SAP has long been saying very similar things, as spelled out at www.dbms2.com/2005/12/09/36/. IBM is also, at least in part, favorably inclined. IBM, however, is in no way promising a comprehensive or widely adopted services repository in any kind of short time frame.
I'll conclude with some quick updates on the broad variety of focused data-management technologies. I've posted more detail on each of these subjects at www.dbms2.com.
- In my January column, "XML Storage: Oracle Should Be Hearing Footsteps," I praised the potential of native XML data storage. News from Microsoft and IBM tests is somewhat mixed. Both vendors stick by their "schema flexibility" stories, and they cite success in complex data interchange for the insurance industry. But the snazzy derivatives-trading app for IBM's Viper hasn't had the performance kinks worked out yet. Meanwhile, text/XML specialist Mark Logic Corp. is working hard to expand the custom publishing market.
- Back in relational land, MySQL has announced a flurry of new transactional engines. A particularly interesting one comes from hybrid memory-centric embedded DBMS vendor Solid Information Technology. And each time I ask, SAP tells me good things about MySQL's progress toward SAP certification.
- Data warehouse appliances continue to show strong early growth. DATAllegro has joined Netezza in the club of appliance vendors with actual announced customers. Interestingly, DATAllegro's and Netezza's products rely on open-source DBMS technologies PostgreSQL and Ingres, respectively.
- In November 2005, "Managing Data at RAM Speed" pointed out the huge potential of memory-centric data management, in online transaction processing and analysis alike. I have expanded that into a white paper, available at www.monash.com/MCDM.pdf, with a lot of added detail on both the software and hardware sides.
- Cogito adds a new wrinkle to data management: a graph-theoretic model that's well suited for relationship analytics, a.k.a. social network analysis.
- As for InterSystems' Cache, the impressively performing object-oriented DBMS that was the focus of my May 2005 column, "Looking Beyond the Big Three," I have little new to report. I actually visited the company earlier this month, but the briefing didn't go well at all. In fact, the entire building had to be evacuated because of a fire ... which turned out to be in the engine of my car; www.monashreport.com/2006/05/13/burning-issues-in-an-analysts-life/ tells the tale.
Curt A. Monash is a consultant in Acton, Mass. You can reach him at curtmonash@monash.com.