BI revisits the mainframe

UK: Mainframes are the legacy platforms of yesteryear when it comes to business intelligence (BI). Or are they? Cognos announced last month that its BI software will be available on the IBM System z mainframe running Linux.

Mainframes are dead; long live the mainframe

"Not so. Trusty mainframes are alive and kicking and remain critical components of a company's IT infrastructure, especially for mission-critical high-volume transactional environments, like financial services, where the mainframe has proven itself to be a trusted platform for housing large amounts of data in a secure and centrally managed environment."

"Research shows that mainframe revenues are rising and MIPS capacity is at an all-time high. We suspect that much of this growth is coming from existing mainframe users who are either upgrading or growing their mainframe usage. At the same time vendors are becoming sensitive to the cost and introducing new pricing strategies to make it more cost-effective on both the hardware and software fronts that extend its reach to companies that once felt priced out. The resurgence in interest in mainframe is also partly due to companies moving to consolidate their IT infrastructures. Mainframe systems like IBM's System z, with its large capacity, are good consolidation tools."

"Since BI applications are growing in terms of data volume and performance, these are points not lost on vendors like IBM/Cognos and SAS Institute, both of whom continue to invest significant development dollars on this platform. For mainframe vendors in it for the long haul, like IBM, the challenge is to sell the mainframe into new customers/applications. BI can help them to do that."

"However, many BI vendors have simply given up the ghost when it comes to extending their reach to the mainframe. By doing so we believe they're missing out on an opportunity. They should think again. We believe there are several benefits to be had by putting BI on the mainframe."

"First, it enables enterprises to confidently (and smoothly) scale up the performance of sophisticated data analysis and other BI functions against larger volumes of data."

"Second, since more customers view BI as a mission-critical application, why not run it on a resilient and scalable mission-critical platform, which plays directly to the strengths of the mainframe, namely industrial-strength processing power, high-availability, reliability, security and centralized IT manageability."

"Finally, BI is still a growing market. Pushing the software onto the mainframe helps companies to both protect and leverage their mainframe investments (i.e., using BI to drive legacy modernization without replacement of the mainframe)."

Debunking the urban myths

"In considering a business case for moving BI data and applications (back) onto the mainframe, customers need to carefully consider the following issues."

"Classic mainframe myths still exist today. Mainframes have traditionally been associated with high total cost of ownership; lack of advanced applications; inability to support realtime/low-latency processing; poor back-end data integration support; shortage of mainframe skills; and steep and inflexible development and maintenance curves."

"While some of that might still be true, most of these sticking points have been, or are in the process of being, addressed. For example, mainframes have certainly evolved dramatically from their early days, when they ran on single proprietary operating systems and single arithmetic processing. Today's modern mainframes have been fitted with new partitioning, virtualization and workload management techniques that host multiple operating systems, can emulate other hardware platforms and have the capacity to support mixed BI workloads without stringing queries in parallel across server nodes that are complex to fine-tune and administer."

"Cost has always been a barrier for mainframe computing, particularly the long-term economic implications of moving escalating and increasingly complex BI requirements. However, vendors like SAS and IBM are now introducing new pricing strategies and open source deployment options that make it more affordable."

"That said, some challenges continue. For starters, mainframes no longer enjoy a wall-to-wall environment like they did back in the 1970s or early 80s. Today, they have to co-exist in a heterogeneous client-server, web SOA, software as a service and increasingly cloud computing world. Can they ever become a good citizen of the corporate IT infrastructure? Customers need to ask themselves how easily they blend into these modern architectures. Our experience shows that there's much work to be done here."

"Even though data volumes and workload processes are spiking, mainframe data center staffing levels have not changed significantly, despite these increases. That perhaps remains the greatest challenge for the mainframe. There are simply not enough young, bright people wanting to learn mainframe skills over PHP, Java, Flash, and other 'hip' Web 2.0 technologies."

The author is senior analyst at Ovum.