A Reality Check

Some people who read this will probably think Doesn't this guy understand that VM has declined, and other types of computing are dominant? Of course I do. As much as I think VM is great, I fully recognise that it is a minor player, with a tenuous existence.

In fact, VM has only a part of what I do for a living since 1995. I have a SPARC running Solaris and a Pentium running Windows NT on my desk, and am responsible for services running these operating systems. I spend most of my time on initiatives that have nothing to do with VM or any other IBM product. I have a LAN at home with NT, Windows 9x, and Solaris. I've even been accused of being a "Unix lover" because of my insistence that our network infrastructure should be "open" rather than Microsoft-oriented. How ironic!

For full disclosure: I'm a shareholder in Sun Microsystems, Microsoft, and IBM, so I put my money where my mouth is.

The attention of computing has mostly gone elsewhere. The desktop today is basically a Microsoft Windows world, with far smaller populations on Unix, Macintosh, and a few others. This is obvious.

In the institutional world you see VM systems (and VAX, and MVS) in use, but they are a smaller and more beleaguered community than in the '70s and '80s. Oddly, Unix is under attack from below by increasingly more functional (and market dominating Windows NT) and from above by increasingly cost-effective and flexibible mainframes. It's also a prisoner of its own track record of expensive failures. It's amusing to see Unix die-hards complain about "toy PCs" that "aren't robust" and "don't scale", sounding just like mainframe-only bigots ten years ago. If Unix doesn't differentiate itself from NT real soon, it's going to have the same crisis that mainframe systems had in the early 1990s. However, major Unix vendors like Sun Microsystems are very aware of this risk, and are putting a smart and credible campaign to emphasize how much more scalable and robust Unix servers are than PCs. (IBM is hoping that everyone will "come back to mainframes", but they're obsessed with OS/390. They don't understand that VM is much more like the Unix style of computing than OS/390 ever can be, and represents a far more natural "upsizing" platform for Unix or NT applications.)

I also acknowledge and benefit from the great contributions made by other systems. MVS was the only venue for large-scale production computing, and still represents the "gold standard" for big batch and transaction processing. PCs and Windows brought a vast audience to computing that was never reached by mainframe and minicomputer systems. Unix was and remains the great innovator that influenced every other operating system environment: it introduced or popularized tree-structured filesystems, distributed filesystems, the pipe concept, command-line multi-tasking, and the TCP/IP communications protocol.

I have no problem praising other systems for their achievements. I just want VM to get its share.

VM influence on computing

VM, though largely unrecognised, has deeply influenced computing. It popularized personal computing with user empowerment and subsecond response times. It invented the client/server paradigm via message-based applications (between different "machines") with fast interprocess communication.

It's surely no accident that both CMS and MS-DOS have A-disks, B-disks, and C-disks as part of a file identifier, that both have EDIT, ERASE, COPY, FORMAT, RENAME, TYPE and PRINT commands that do essentially the same things. In Novell and Windows NT environments, you also see commands with functions and names very similar to TELL and ACCESS. I also have to believe that CMS's PROFILE EXEC inspired Unix's .profile and .cshrc, or MS-DOS's AUTOEXEC.BAT. There are EXEC2 programming hacks that have almost identical analogues in Unix shell programming, sometimes with only minor syntactical differences. This can't be accidental either. It is noteworthy, though, that the EXEC2 tricks are folklore, but the shell equivalents are still in use. Sad, no?

The virtual machine concept has been widely adapted on other computing systems as a valuable abstraction: one cannot read about Java without repeated references to the Java Virtual Machine; the same jargon is used with Smalltalk. VM's organization, with some services provided by the kernel and others by privileged virtual machine processes, clearly influenced the design of modern operating systems, from Mach onwards. Wouldn't it be great if IBM printed an advertisement that bragged about having pioneered this great concept?

Despite the changes in the computing world, VM still has a great deal to offer, both as an example, and as a working, running system with high value.


[ Prior page | Next page | Top page ]