With the unveiling of new hardware, 2008 was a busy year for the mainframe. Armonk, N.Y.-based IBM Corp. released the Enterprise Class and its smaller sibling, the Business Class. But that wasn't the only mainframe news of the year. Software vendors attempted to ease software costs for mainframers, Linux on the mainframe continued to grow traction, and competing vendors made their best pitch for migration off the platform.
The Mainframe: Month by month
January. We opened the year with some mainframe New Year's resolutions. Tipster Robert Crawford cautioned against outsourcing too much and warned that independent software vendors (ISVs) have helped kill the mainframe. According to Crawford, IBM has done well in enabling big iron for the Web, but it needs to get back to its roots and fine-tune foundational firmware like CICS.
Speaking of ISVs, mainframe software giant CA announced a plan to reduce mainframers' software costs by offering the option to buy according to millions of service units (MSUs) instead of millions of instructions per second (MIPS) on the reasoning that MIPS doesn't measure the actual consumption of work, while MSU does.
February. February was the month of the z10 mainframe. SearchDataCenter.com obtained an internal IBM document on the new mainframe and broke the details a week before the official IBM announcement of z10. The new big iron has the potential for about 1.5 TB of memory and 64 4.4 GHz processors, about 100% more CPU performance, and triple the memory capacity of its predecessor, the z9.
Novell also made a splash by offering a free starter kit to run SUSE Linux on System z to facilitate the transition for those migrating to zLinux.
Meanwhile, Crawford busted five mainframe myths, including that the mainframe is old technology, that it requires too many human resources to manage, and that big iron can't compete in an Internet-driven world.
March. In March featured news and stories from the Share user group conference, which took place in Orlando, Fla. George Handera, a systems engineer for a major insurance company and a beta user of the z10 mainframe, said that HiperDispatch has brought promising mainframe performance improvements.
Also in user news from the show, Jim Horne from Lowe's discussed how determining mainframe capacity isn't what it may seem because MSU can be imprecise. Horne recommended the Java-based tool zPCR, or System z Processor Capacity Reference, to get more accurate capacity data.
Also at Share, Mark Post from Novell compared data center buildouts using Intel-based Linux servers with a mainframe architecture. His conclusion? An Intel-based architecture is about 66% more expensive.
April. At the AFCOM user group show in Orlando in early April, SearchDataCenter.com spoke with Ralph Crosby, the chief technology officer of BMC Software's's mainframe business. Crosby discussed virtual machine sprawl on the mainframe and how the z10 has boosted WebSphere performance, among other topics.
Also on the topic of the z10, tipster Wayne Kernochan discussed using the z10 as a data center hub.
Later in the month, we asked whether SOA and BRIC – which refers to the sales markets of Brazil, Russia, India and China – had depleted mainframe innovation.
May. We got an update on the New York Stock Exchange (NYSE), which last year began migrating off the mainframe and, at the end of April, unplugged its last big iron unit. Francis Feldman from NYSE's technology division said that although the migration made sense, it was still cause for mourning, as he is "an old mainframe hack."
Kernochan wrote about the mainframe as outmoded when it comes to open source applications. In particular, he stressed that the open source community has moved well beyond Linux, while the mainframe hasn't.
We also featured a lively blog discussion on the cost savings of running mainframe specialty processors such as the Integrated Facility for Linux (IFL) and the z Integrated Information Processor (zIIP).
Adding some levity, Crawford offered up some practical jokes for mainframe systems programmers.
June. Piggybacking off the previous month's discussion of specialty processors, a director in the IT center in Harris County, Texas – the third largest county in the country – said that determining how much eligible work can fit on a zIIP or a z Application Assist Processor (zAAP) can be a "dark art."
I followed up my post on specialty processor benefits with another, "Indirect vs. direct savings on the zIIP and zAAP: What's the difference?"
On the tips side, Kernochan discussed virtualization and its reduction in the total cost of ownership for a z10. And Crawford argued that when it comes to mainframe vulnerabilities, it's best to be proactive.
July. IBM announced its purchase of Platform Solutions Inc. (PSI), a plug-compatible mainframe startup with which IBM had been in legal wrangling for more than a year. The acquisition brought further details and speculation, and many competitors complained that IBM bought PSI to put them out of business.
Robert Crawford argued that it's time to commoditize mainframe software costs. "Personally, nothing is more aggravating than having to pay more for the very same software just because the new CPU runs faster or the box has more processing capacity," he wrote. Many mainframers agree.
Also in July, IBM announced that it would enhance the functionality of z/VM on the z10 mainframe by allowing it to use multiple specialty processors (IFL/zIIP/zAAP) within a single logical partition (LPAR). Later in the month, Jeff Savit from Sun and some IBMers got into an online fracas over the virtualization benefits of the z10. We deemed it a cage match.
Matt Stansberry, SearchDataCenter.com's senior site editor wrote a detailed piece on zLinux. In particular, Stansberry addressed the speculation that more than half of the mainframe capacity that IBM sells – measured in MIPS – were to run Linux on the mainframe.
A big hit was a Share user group conference video from San Jose, Calif., of a group of mainframers on a morning jog. Mainframers may be getting old, but they're doing their best to stay healthy, – or at least some are.
Other news from Share included these features:
- The conference keynote, which focused on virtualization, energy consumption and cloud computing, and not the mainframe.
- A list of seven things all mainframers should know
- A video interview of a mainframe student anticipating a bright future
September. It was a month of tips. Crawford wrote about four ways to add flexibility to mainframe applications, as well as essential tools for mainframe operation and management.
Meanwhile, Kernochan wrote about the future of the mainframe in the health-care industry, and using the mainframe as a "green monitor."
Crawford also blogged about the effect of several historic IT trends on mainframes, including structured programming, personal computing and Java.
October. IBM pushed out its z10 Business Class, the smaller sibling of the z10 Enterprise Class it announced earlier in the year. As part of the announcement, IBM offered what amounted to a layaway program for the z10 Business Class for shops that want their big iron now but might not have the budget in the fourth fiscal quarter of this year.
OpenSolaris on the mainframe became available and will run on an IFL under z/VM, similar to Linux. We spoke to Sine Nomine Associates' David Boyes, who developed the technology.
Finally, Andy Baritchi, a senior security consultant for a Fortune 100 company, opined that cloud computing was just a mainframe rehash.
November. We explored the possibility of deploying a Linux-only mainframe, compared with running big iron that has Linux and z/OS on the same box. Incidentally, there are pros to both approaches.
Crawford discussed the oft-mentioned new feature of the z10 HiperDispatch. In particular, he explored using HiperDispatch for vertical CPU management.
Finally, thanks to a report and release from analyst firm Robert Frances Group, HP and IBM got into a back-and-forth. HP announced that over the past three years, 250 customers had migrated from the mainframe to its Itanium systems and cited an RFG study in the process. RFG took umbrage at a paragraph and sent out a press release stating that the mainframe is often the best server hardware platform choice.