NOSes /

Microsoft's newest nemesis - the Linux mainframe?

By Kevin Tolly
Network World, 03/04/02

Years ago, Microsoft's Windows NT Server vanquished IBM's technically superior OS/2 Warp in the struggle to own the then-nascent realm of enterprise servers.

Since then, Microsoft largely has focused on trying to gain market share in the high-end server space from Sun. Now with the release of the new IBM zSeries 800 e.server (aka z800), Microsoft and Sun may have a lot to worry about.

The machine clearly is targeted at the high-end enterprise and e-business data center application arena these two treasure - and claim to own.

IBM says that a single z800 can consolidate the workload of a data center contingent of eight Sun 3800s and 12 Compaq DL580s, thus taking on the kind of load you'd expect from a mainframe.

It has a lot going for it. First and foremost, unlike other servers, it is a mainframe. IBM's advertising copywriter makes certain the reader gets that point in a recent ad in the Wall Street Journal: "The z800 isn't like a mainframe, it is a mainframe." That means a lot.

Anyone who has ever implemented and supported an IBM mainframe data center (as I have) no doubt has cringed over the years as Microsoft touted its NT server environment (and Windows 2000 is just NT Version 5) as nonstop, reliable platforms for mission-critical applications.

There is simply no way that a server, essentially a souped-up PC, can hold a candle to the IBM mainframe when it comes to hardware reliability and redundancy.

While my firsthand knowledge of mainframe hardware is getting a bit dated, I'd bet 10-year-old mainframe technology still outclasses much of what is touted as high-end, highly reliable server hardware.

While "better, faster, cheaper" mainframes have been around for a long time, the clear operating system of choice was IBM's high-end MVS. Even with inexpensive hardware, licensing and system software complexity made it ill-suited for the "server farm."

And, while IBM ultimately embraced open protocols such as TCP/IP and delivered mainframe-based applications FTP and Web services, they were not typically stellar performers. They were written to IBM's MVS operating system, and its APIs reeked of "proprietary" - something nobody wants if it can be avoided.

What's different now is that this mainframe offers native support for Linux. With that move, IBM solves multiple problems.

For starters, it gets an instant application suite. Along with Web services, file storage and transfer, and e-mail, the myriad Linux-based applications from third-party developers are candidates for deployment.

IBM's system programmers can focus on the underlying Linux/mainframe interfaces to maximize performance and reliability. Its application programmers can pick "sweet spots" and not worry about providing every application imaginable.

Gone too, along with the burden of steep software licenses, will be the "army of rocket scientists" needed to maintain the system - or so reads IBM's advertisement. And, while Linux isn't as easy to use as Windows 2000, it is a darn sight easier to use than MVS.

Not only will the ranks of people capable of running Linux environments grow over time, but also Linux itself will become easier to use as it matures.

For several years now, I've heard complaints about the lack of server scalability from network managers of large enterprise and e-business data centers.

Outfitting, maintaining and managing dozens upon dozens of servers that are all doing pieces of the same job - such as Web services - is an inferior solution to a serious problem.

If IBM can deliver anything close to what it claims, network managers might finally get a viable alternative to the server explosion of recent years.