The Mainframe's Future is Analytically Cloud-y
I have blogged several times about IBM's renewed (and bigger) push into analytics and business intelligence, and so has James Governor recently. IBM has completely changed its strategy, pushing aggressively to build and to promote business intelligence-related solutions on System z, especially those concerning "operational business intelligence" where the mainframe has unique advantages. Customers have been pleading with IBM for years to take another look at BI on z, and IBM is stepping forward (finally) with some new and big ideas.
I remember talking with a CTO at a major financial services company years ago about his problems meeting his end users' requirements for more and better business reporting. There was (and is) an ever-growing need for more and more timely analytical information. He pointed out that his two mainframes literally ran almost the entire core business efficiently and cost-effectively, processing all the core transactions and data for his firm and scores of others. (Only two machines! This company also served several other financial institutions, offering common services. True cloud computing before the term was invented.) But, for reporting, it was no longer viable for his development team to write "yet another custom program." They wanted to take a more streamlined, standardized, end user-driven approach that could respond much more quickly to ad hoc business requirements.
At the time he asked IBM for a solution, and they said, no problem, install additional servers — probably hundreds, to meet regulatory requirements for financial data security — start with a big ETL (Extract/Transform/Load) each night, run your analytics (on stale data, no doubt), and then push some of the (stale) results back to your operational databases on the mainframe. And the CTO, to his credit, said (closely paraphrasing) "Are you bulls**tting me? Go back to the drawing board. You're no better than Oracle." And he wasn't the only CTO saying much the same, often quite forcefully.
So IBM went back to the drawing board, and how. IBM quickly realized that there's no other feasible way to solve many major analytical and business intelligence problems without placing those functions directly in and on the operational systems. And that quite often means mainframe hosting, of course. Ergo, IBM has invested heavily to imbue the mainframe with new BI-related capabilities: Cognos, SAFR, ILOG, dashboards, event-driven middleware, Smart Analytics Cloud for System z, etc. (More to come, I'm sure.) IBM has also introduced solution packages so customers can say, "I'll have one of those, please" and IBM can quickly present a complete proposal at a competitive price.
IBM is really onto something, I think. I am hearing from so many customers about their pressing, unsolved BI and analytical problems — and how if they can solve these problems they would gain tremendous competitive advantage. I'm also hearing from people who manage IT infrastructure who are frustrated with the current too-high cost deployment patterns for these solutions. A lot of them are seeing massive and growing server farms that run busy for only a couple hours per night, per week, or even per quarter. That sort of deployment is unsustainable from a cost point of view, so managers are certainly looking at virtualization and cloud computing. But one problem is security: the information these systems handle is often extremely sensitive, and only the mainframe (with its unique technologies and security certifications) is capable enough to virtualize many of these sensitive data workloads. I am also hearing increasing problems with the ETL step: it's getting more and more difficult to physically move data through the ETL stage(s) as operational data grows. Why not improve the efficiency of ETL by bringing data warehouses physically closest to operational databases? System z does that best. Better yet, how about avoiding ETL completely, at least for certain tasks?
I am seeing high and growing demand for System z-based BI solutions. Many customers already "get it." The global economic environment is causing many others to reconsider their architectural approaches, to incorporate the mainframe more strategically (and more directly) into their BI/analytic service delivery. I think IBM is quite smart to invest in this area, and I'm pleased to see IBM's new thinking and its results.
|by Timothy Sipples||October 6, 2009 in Innovation |
TrackBack URL for this entry:
Listed below are links to weblogs that reference The Mainframe's Future is Analytically Cloud-y:
Thanks for very informative and very helpful article.The cloud concept is very interesting in web era and I think the mainframe is getting more in cloud concept.
Posted by: bluetooth freisprecheinrichtung | Oct 10, 2009 6:44:27 AM
The comments to this entry are closed.