z/Journal: Over the last 15 years, competitors’ propaganda has successfully cast the mainframe as being outdated. Will we see anything from IBM in the next few years to improve the overall image of the mainframe?
Jim Porell: No need to wait a few years, we’ve already started! There are two fundamental areas we are constantly working on at IBM. First is continuing to grow our traditional business through large-scale, better availability, etc. But at the same time, improve the consumability of the platform. What does that mean? Make it simpler. Improve on defaults. Take a multi-step command line work flow and make it a point-and-click operation. Make some of the administrative work consistent with other platforms to facilitate greater skills portability. The second area of focus is to provide new levels of integration between our traditional or heritage workloads and new workloads, such as Web Services and Service-Oriented Architecture. And as we do that, we also need to change the economics. We’re doing that through new processors, such as zAAPs (zSeries Application Assist Processors), zIIPs (z9 Integrated Information Processors) and IFL (Integrated Facility for Linux) as well as leveraging accelerators, such as our recent DataPower acquisition, to simplify XML and SOAP processing.
z/J: Is there anything in the works that will lower the barriers to entry at the low end? If you’re a small business, consumability includes being able to afford it.
JP: The z890 processor included an ability to ratchet down the speed (number of MIPS) in the general purpose processor, while providing full-size engines as zAAPs and IFLs within the server to capture new workloads. You can expect to see a similar approach for the next generation of low-end processors, while adding in the new engine types such as the zIIP, to continue to change the economics for processing work with z/OS. And then again, how small do you have to go? Well, Small- and Medium-Size Businesses (SMBs) can go to Strategic Outsourcers, such as IBM Global Services. They can host the workload for a small business and provide the value of the mainframe under the covers. The customer isn’t burdened with managing processor capacity and IBM System z, but gets the benefit of the years of architecture and integration that IBM has built in to make it a very efficient transaction and database server. Our services folks can provide very big things in small packages.
z/J: For CIOs who come from Wintel and Unix backgrounds, would you tell them why specialty processors are a big deal?
JP: z/OS is all about integration, which in turn reduces operational complexity. CIOs who are experiencing the pain of scalability limitations and/ or server-farm proliferation are in the best position to appreciate this approach. The ability to have multiple applications, middleware and databases on the same operating system image can dramatically reduce the number of physical servers necessary to meet the demands of an enterprise. For example, grocery chain Hannaford Bros. is replacing 500 Intel-based servers with one virtualized z9 running Linux.
We’re seeing a groundswell of interest centered around how mainframes address a number of thorny corporate issues: securing data, business resiliency, complying with regulations, and more. Java, Linux and SOA are making the mainframe more attractive, and specialty processors give us yet another dimension for optimization. One large retailer recently proved that a Java application from a competitor’s Unix server delivered eight times more throughput running on an IFL.
z/J: For the last 15 years, such things as security, resiliency, and workload management continue to define the mainframe as the industry’s premier server. Why do alleged mainframe alternatives still lag so far behind in so many functional areas?
JP: There are five areas of strength in the mainframe: security, workload management, business resilience, business process integration, and storage management. I like to think of enterprise computing as the food pyramid: mainframes, with the fewest systems at the top, above Unix and RISC systems, with more servers and they are above Intel systems, serving Linux and Windows with the greatest volumes. Each of these levels manages downward: Windows does a great job of taking care of itself, but not so great a job managing other platforms. Unix systems manage themselves and Intel systems pretty well, but don’t see enough opportunity in managing those “niche” mainframes. Mainframes can’t really participate in the enterprise by themselves. Why? Because for some reason, the punch card and 3270 green-screen terminal didn’t become as ubiquitous today as we anticipated! Well, seriously, our model is to embrace and integrate other platforms. So, we look at our five core strengths for the mainframe, but also as an enterprisewide role. We look at how we can make the mainframe more available, while at the same time, making an end-to-end workflow more available. How? By leveraging capabilities such as Global Mirroring or XRC within the storage controller. We can attach a non-z/OS disk to the system and make volume copies. So, if there’s a disaster, we have our data copied as well as theirs. It’s much faster to restart their servers and re-connect to their data than worry about finding tapes and re-loading servers. We’ll introduce that capability on Linux for System z first, and add it to other platforms as we go.
Business process integration is probably the biggest thing going for the mainframe now. People don’t realize we’ve been doing that for many years. In essence, running CICS and IMS, alongside VSAM, DB2 and other vendor products, has been a level of integration for years, and one that’s backed by a system integrity statement that works for a business. Now, adding Web Services to the mainframe is just another piece the enterprise puzzle. Our integration can drastically simplify the operational model as compared to multiple systems being networked together to accomplish the same workflow.