Trends > z-OS

Perfect Match

BI solutions give mainframe customers an edge, IBM’s Karl Freund says

At an IBM Design Center in Poughkeepsie, N.Y., database architects shape the future of business intelligence (BI)—the near future. The focus is on the near future because the limits of current technology are continually being pushed. Customers are accumulating data at a dizzying rate, and these days, they require almost instantaneous, organization-wide access to all flavors of mission-crucial BI.

“We have to be ahead of the curve,” says Caryn Meyers, IBM worldwide marketing manager for BI solutions on the System z* platform. “We can’t wait for customers to get there.” As part of its efforts to stay a step ahead, the Poughkeepsie team has been looking into what the mainframe can deliver as an enterprise-scale BI platform. In the summer and fall of 2008, they put DB2* for z/OS* to the test to see how it managed large data volumes and queries for enterprise-level implementations.

For what was called the 50-Terabyte Study, team members created—you guessed it—a database of approximately 50 TB, and hosted it on an IBM System z server with about 250 TB of disk. It’s impressive indeed, but hardly unprecedented; a handful of enterprises actively run databases that size or larger. The cool thing with the 50-Terabyte Study is most of the IBM-created data—about 35 TB—resided on a single table. Now that’s something you don’t see everyday–not just yet, anyway.

Today’s large databases consist of thousands of tables, with information fragmented across these environments. Database administrators have come to rely on this design because, unwieldy as multitudes of tables may seem, it’s been the best available and most efficient option. The alternative—pouring all this data into one or a few large tables—would generally require greater system resources be devoted to essential tasks like running queries and utilities.

“It’s certainly simpler to manage and control one table,” says Meyers. “Is this necessarily what every customer would want to do? No, probably not, because not every customer will want to scan the data. But customers are growing their data, and it looks like they’re heading in this direction. We wanted to make sure we’re in safe water, that there weren’t any surprises.”

By “no surprises,” Meyers means the whole process of loading, scanning and managing this enterprise-warehouse environment was a success. When tasked with creating a table of about 300 billion rows, DB2 code was executed. Workload Manager efficiently prioritized the individual queries, with the 50-some engines supporting this environment providing performance and scalability.

In late 2008, IBM continued to examine the results of the study. The company is expected to release detailed findings early this year, but throughout last year, word about this mammoth table reached select customers. “Big customers come in and talk to us,” says Meyers. “They’re very interested in hearing about this.”

To be sure, the results of the study are significant. But for all of the terabytes involved, it’s only a small part of a larger story. The larger story is as BI has evolved from a departmental luxury to an enterprise-wide priority, it’s been embraced by the ever-adaptable mainframe. It turns out the mainframe and BI are a natural fit.

BI Grows Up

Data is now a precious business commodity. Information must be secured, and it must be available in real-time. Availability and security are the mainframe’s hallmarks. So why wouldn’t you want the mainframe to host your warehousing environment? Go back 20 years, or even a few years, and many of the most knowledgeable and loyal customers didn’t view the mainframe as a system well-suited to hosting BI applications. At the time, they were right.

Initially, business analytics was generally limited within the enterprise. Maybe one department head would scour data to conduct analysis or churn out reports. Information’s currency wasn’t a concern, since analysts were just beginning to tap into BI’s potential to deliver real business solutions. It was a small operation, and the mainframes of the time weren’t designed to perform computationally intensive tasks. It wasn’t uncommon for these individual departments to address their BI needs by bringing in a low-cost, higher-performing UNIX* system and attaching some storage to it.

Of course, times have changed. The BI bandwagon is rolling. Customers retain virtually every byte of the now-massive amounts of data they gather, knowing this information, once carefully analyzed, could yield transformative business solutions. Data extraction and analysis occurs in real-time, in every corner of the enterprise. Old solutions, practical and cheap no longer, have been outgrown.

“The problem, of course, is you’re spending perhaps millions of dollars on a data warehouse that replicates data just sitting over on the mainframe, on which you’re actually conducting the transactions,” notes Karl Freund, IBM vice president of strategy and marketing for the System z platform.

Meyers adds, “Availability of information is everything. Conducting a summary tonight or at the end of the week is no longer an option. Now it’s becoming really important to deliver to the front lines of business—to the call centers, to the managers who make immediate decisions, to the operation managers who run plants. It’s becoming much more of an immediate type of decision-making that’s being integrated into operational processes. So customers are rethinking how they structure their BI environments.”

Availability of information is everything. Conducting a summary tonight or at the end of the week is no longer an option. Caryn Meyers, IBM worldwide marketing manager, BI solutions on the System z platform

Next page: >>