If META is right, a whole bunch of people at IBM and mainframe Linux distributors SuSE, TurboLinux, and Red Hat should get ready to post their resumes on Monster.com.
The META report says that mainframe configurations of Linux will fall out of favor as soon as 2005, by which time Unix and Intel-based Windows and Linux solutions will have enough mainframe-like features without the mainframe-like costs to make those options the better choice. In a scathing review of Linux's ability to support mission-critical applications, the report cautions that "current Linux incarnations are relatively immature, as evidenced by the interminable list of errors/patches on Linux providers' Web sites. They have, therefore, been naturally limited to simple, less-critical applications such as LL Bean's e-mail notification system."
Giga senior analyst David Mastrobattista thinks the prediction is a bit premature: "The META article was overly negative for a technology that is just beginning to see the light of mainstream use." According to Mastrobattista, more than 300 mainframe sites are doing something serious with Linux already and that number is going up. Referring to recent entries from Computer Associates, SAP, Oracle, Candle and BMC, he says, "it's beyond the first phase of early adoption at this point. Now that applications that support Linux on the mainframe are starting to surface, it is making inroads into serious financial companies on Wall Street, pharmaceutical companies, and insurance companies." Mastrobattista says that mySAP.com is the first real big example and that it will begin shipping later this month or in early June.
IBM vexed by report
It should come as no surprise that IBM --- the biggest target of the report --- is also vexed by META's findings. IBM's director of eServer zSeries product marketing Peter McCaffrey has no qualms with META's findings that most mainframe Linux implementations so far have been limited to less mission-critical applications such as e-mail servers, file and print sharing, and Web serving. According to McCaffrey, it's precisely those infrastructure applications that are sitting on individual servers sucking up valuable corporate resources. "With a mainframe model," says McCaffrey, "You can virtualize all of those systems onto one mainframe. When an application on one of those virtual systems is idle, another application can use that capacity."
But the META report doesn't take issue with the promise of server consolidation. It takes issue with consolidating them using mainframe Linux. By 2005, the report says, lower cost Unix and Intel-based platforms will inherit some of the availability and workload management features that make mainframes today's most appealing consolidation targets. It's a valid point, considering that today's total cost for an entry-level mainframe running Linux costs in the neighbourhood of US$500,000.
But you don't have to throw a stone too far before you'll hit a bunch of people who will roll their eyes every time they hear another "mainframe is dead because of the closing gap" theory. "How many times have we heard this before?" McCaffrey asks. "The last time we heard this was with the client/server revolution. Meanwhile, in the last 18 months, IBM has sold just over 100 brand new mainframes just to run Linux. With more enterprise applications coming on-line, the interest is rising."
Giga's Mastrobattista also is lukewarm to the notion that, by 2005-2007, the mainframe will lose significant ground to lower-end systems bearing mainframe-like prowess. "If you rate capabilities like workload management on the mainframe as a 10, it will take until 2005 and 2006 for those systems to start catching up and even then, it will only be an 8." says Mastrobattista. "And it won't be until the end of the decade before those systems get any real traction in the market."
IBM's McCaffrey also believes that the META report assumes that mainframe technology is standing still. "We've made significant investments and we will continue to raise the bar. I don't see Wintel catching up in this time frame or any time frame." If IBM were only a vendor of mainframes, I would take that last statement with a grain of salt. But IBM is one of the premier vendors of Intel-based servers too. The company probably has a good idea of where the gaps are and whether they are widening or narrowing.
In addition to the research and development that IBM is pouring into its mainframe offerings, the old staples of mainframe-class systems (like their legendary I/O throughput) will not be threatened by any up and comers any time soon. Giga's Mastrobattista says, "I/O has always been a big strength of the mainframe. When it comes to commercial, mission-critical database processing, there's no platform that can match a mainframe."
What's my take? Well, I've spent the last six months studying mainframe Linux, interviewing CEOs like Sun's Scott McNealy and Intel's Craig Barrett, and digesting reader feedback to all the coverage. There's no doubt that the low-end is looking more and more like the high-end. In fact, maybe the question we should be asking is "what the heck is a mainframe anyway?" Seems like tomorrow's Intel boxes are looking more like yesterday's mainframes. Nevertheless, META's prediction is a bold one-- perhaps a bit too bold for me. And you know me: I normally like to go out on limbs.
So what's your take? Tell talkback whether Linux has a future on mainframes?
Another article predicting Linux's downfall...