|Editions | myCNN | Video | Audio | Headline News Brief | Feedback||
Linux looms on the mainframe horizon
(IDG) -- To the Unix community, it is now merely the rumble of distant thunder on an otherwise sunny day, but rising corporate interest in Linux-based mainframes could have them reaching for an umbrella in the next year or two.
Driving this heightened interest in mainframe Linux are the spiraling costs associated with buying and managing the mushrooming number of departmental Unix-and Windows NT-based servers. During the past few years, these systems took over the running of key Web-based applications now central to corporate users' competitive strategies.
In the past three months, top-tier Linux distributors including Red Hat Inc., TurboLinux Inc., and SuSE Linux AG have pledged or deepened commitments to work with IBM Corp. and deliver native versions of Linux that exploit Big Blue's mainframes. IBM will underscore that relationship by bundling Linux versions of key middleware products, such as WebSphere, MQSeries, and CICS, with the new versions of the operating system.
What is also inspiring corporate interest is this month's release of the long-awaited Version 2.4.0 of the Linux kernel, which features several critical high-end server capabilities, including support for IBM's S/390 and z900 mainframes. IBM Chairman Lou Gerstner also announced in December that IBM will invest US$1 billion in Linux during 2001, a good portion of which will go toward the development, marketing, and servicing of Linux on mainframes and supercomputer-class clusters.
Big Blue's streak
Evidence that corporate users are more than curious about the specter of Linux mainframes is found in the handful of recent customer wins scored by IBM. In December, Telia, Scandinavia's largest telecommunications company and ISP, installed a Linux-based IBM G6 mainframe to replace 35 Sun-based servers used largely to run its billing system.
Meanwhile, Grede Foundries, a $600-million-a-year metals company in Milwaukee, installed Linux on its existing S/390 to handle Web serving, file, and e-mail functions, taking those duties away from a large base of Unix-and Windows-based servers.
"Over the next year, I think you will see a number of wins where IBM convinces Sun Microsystems, Hewlett-Packard, and Compaq shops that their cost of computing can be lowered if they consolidate their midrange systems onto a single [mainframe-based] Parallel Sysplex system," says Dan Kusnetzky, vice president of system software research at IDC, in Framingham, Mass.
By pushing Linux on its largest systems, IBM increases its chance to sell more $1 million mainframes, which would go a long way toward shoring up sagging mainframe revenues. For their part, Linux vendors get to use a Big Blue club to beat down Unix competitors unwilling to work cooperatively.
"Linux on mainframes is strategically important to us because it can divide the server into thousands of virtual servers, each running an independent application. It's also giving [the mainframe] a cooler image in both business and universities, where we are building up a much-needed skills pool among students," adds Pete McCaffrey, IBM's director of enterprise servers, in Poughkeepsie, N.Y.
But Sun and HP executives view Linux mainframes as merely another option for users, one with limited potential to eat into their market share.
"Server consolidation in general is a good thing; it can lower costs and ease management," explains Herb Hinstorff, manager of the Linux and open-source program office at Sun Microsystems, in Palo Alto, Calif. "But server consolidation doesn't necessarily mean moving to a mainframe. Moving to a larger box of the same operating system environment is a better way to consolidate."
Hinstorff notes that Sun is "embracing" Linux by working to make Solaris interoperate with it, but is not planning to sell any hardware with Linux.
"We're seeing just the opposite of consolidation. More people are going to the 1-unit 'rack and stack' clustering environment for Web server reliability," says Mike Balma, HP's director of marketing for Linux systems operations.
Balma points to smaller servers' six-month turnaround time, which allows for a faster upgrade cycle than a mainframe's two-year design cycle as a big reason for that movement. Customers can also add servers easily when needed in one part of operations and reallocate them physically if need be.
There is healthy skepticism among corporate users and suppliers that replacing Unix and Windows-based servers en masse with Linux-based mainframes is in their best economic and technically strategic interests. Doubters cite the lack of a wide range of exploitative applications for the platform, few real-world examples that prove the alleged lower costs of ownership, and, according to one major Linux distributor, the lack of economic opportunity.
"There are a couple of reasons why we aren't going after [the mainframe] market," says Drew Spencer, CTO of Caldera, in Orem, Utah. "One is the [low] volume of those systems, so the money is not enough to justify us going into that space. Second, service and support on these deals, for the most part, is covered by IBM."
Caldera is convinced that the better volume opportunity for both itself and its users is with midrange servers, particularly those expected to emerge later this year around Intel's forthcoming IA-64 chip, according to Spencer. He thinks incorporating Linux as the front end of a three-tier architecture is a better idea all the way around.
Later this month at LinuxWorld Expo in New York, Caldera and the Santa Cruz Operation (SCO) will launch their new company, a melding of Caldera and SCO's server software and professional services division, and lay out the new entity's product plans and strategies. Much of those plans will revolve around clustering technologies and cluster management, centered around Caldera's Volution systems management software, which weaves together existing Unix and Linux systems.
Some pro-mainframe Linux researchers and consultants believe many IT executives overlook or do not accurately assess all the factors involved in running environments with hundreds or even thousands of Unix-and Windows-based servers and often put too much emphasis on the up-front costs of existing vs. new hardware and software.
"In our research, we find that most companies make decisions based on something as simple as what is the lowest-cost hardware platform they can get. But as their businesses evolve into these more complex electronic environments, they discover that sort of thinking is shortsighted," says David Boyes, chief strategist at Ashburn, Va.-based Sine Nomini consultancy and research firm.
For example, Boyes oversaw a TCO (total cost of ownership) comparison study on behalf of a large corporate client, using two very different infrastructures. The first solution involved 750 Sun servers; the second was anchored by a single IBM S/390 mainframe with the company's venerable VM OS governing Linux, which was set up in a partition.
Taking into consideration the cost of hardware, software, middleware, 750 LAN ports to connect the servers, consulting time to set it all up, and the necessary floor space needed to accommodate the servers, the Sun solution was priced at $55 million. The equivalent IBM-Linux solution cost $7 million.
The client also dramatically reduced the time it took to bring new servers up, primarily on the strength of VM's capability to partition Linux, which can then operate thousands of far-flung departmental servers. According to Boyes, it took an average of 1.5 days to set up a Sun server, as opposed to 90 seconds for the mainframe to bring one online.
One factor overlooked by many TCO studies favoring Linux mainframes is the expense of moving various versions of even the same application from one architecture to another. One high-level IT executive at a Fortune 10 company, who requested anonymity, assigned a small team to investigate what sort of technical and financial effort it would take to put Linux on company mainframes. But he will not write off on the project until the applications he is using now are ported over.
"They don't have Netscape Enterprise Server over there [on Linux] yet, and we are not happy with the IBM equivalent," he says. "We looked at it last summer, but it had the sort of cracks and anomalies that make it fall apart in some situations. All I could tell IBM was, 'Give Netscape a bunch of money and have them put it on their box.' "
Linux on desktops: Hurdles remain
Although Linux mainframes may loom on the horizon, Linux on the desktop does not even figure into the forecast.
There have been a few positive signs of late that Linux is making some progress as a meaningful competitor to Microsoft's Windows, but many believe it still has a long uphill walk in the rain ahead.
Sun Microsystems has made the Linux version of its popular Star Office desktop suite available free of charge. Eazel, a promising startup that includes Mac OS designer Andy Herztfeld and former Apple Software evangelist Mike Boitch, is inching closer to delivering its impressive-looking graphical interface that is tied to its networking environment.
Corel's WordPerfect, with a strong following in several key desktop markets including legal and medical fields, has been another shining hope for desktop Linux. But with rumors swirling that the financially troubled company was about to sell off its Linux business, it only further supports the argument that ISVs have trouble making a living from Linux-based desktop applications.
But with less than a 4 percent share of the desktop market at the end of 2000, according to numbers from Framingham, Mass.-based IDC, some are wondering when the open-source OS is going to make any sort of breakthrough.
"I applaud [Eazel's] attempts at making Linux friendlier. But when you get down to it, a pretty interface with no software or missing components on the applications side only means you have a desktop OS that is all dressed up with nothing to run," says Dan Kusnetzky, vice president of system software research at IDC.
The other bad news is that some analysts expect Microsoft's Windows 2000 to pick up steam in both the desktop and server markets this year. With large corporations finishing up their Windows 2000 evaluation and testing cycles and many exploitative applications on the way, many are forecasting much healthier sales even by midyear.
Other analysts believe Linux's best chance may lay with devices rather than desktops. For example, Linux market leader Red Hat, through its purchase of Cygnus in late 1999, is focused on creating embedded versions of Linux for a variety of handheld devices through its EL/IX set of APIs. With EL/IX, developers can create applications on handhelds that require only 32KB of memory.
"We are starting to see the growth of companion devices that sit side-by-side with the desktop systems," Kusnetzky says. "It is looking more likely that this is where Linux will find a place."
Linux stays on the radar screen
The open-source operating system continues to have an impact and influence on large industry players' future plans.
'Ramen' worm hits some Red Hat Linux servers
RELATED IDG.net STORIES:
IBM touts Linux-based supercomputer
|Back to the top||
© 2001 Cable News Network. All Rights Reserved.|
Terms under which this service is provided to you.
Read our privacy guidelines.