By Design

What Is an iSeries?
Article Information
Article ID: 17055

Pub: iSeries NEWS

Dept: By Design

Date: September 01, 2003

 Printer Friendly

Related Topics:
iSeries Issues

Other Articles By:
Frank G. Soltis
Article Feedback:
Submit feedback about this article

by  Frank G. Soltis  

September 01, 2003   Although most of the world's population would have difficulty answering the question "What is an iSeries?" it should be a fairly straightforward question for those of us familiar with this server. Or is it? We may live with an AS/400 or an iSeries every day, but when it comes to defining exactly what it is, many of us may stumble a bit.

It's an IBM eServer. It's a midrange server. It's a business computer. It's used by businesses all over the world. It's sold in more countries than McDonald's hamburgers. It has the largest number of customers of any IBM server. It's all of this and more, but we still don't have a good definition of an iSeries and what makes it unique among servers.

With the movement to converged hardware among IBM servers and the recent announcements that the iSeries will run the AIX operating system in the near future, the definition of an iSeries has become increasingly blurred in some people's eyes. The purpose of this article is to try and clear up some of this blurred vision.

The Problem with Definitions

At the time of the model i890 announcement, I decided to write a whitepaper entitled "What Is a Mainframe?" The model i890 is clearly a mainframe-class server, and at the time we were thinking about simply calling it a mainframe. But you may have noticed that today IBM does not call the model i890 a mainframe, at least not in any official publications. The reason is that within IBM we could not agree on what was a mainframe.

On the surface, it seems that finding a definition for a mainframe should have been easy. After all, IBM has been building mainframes for almost 40 years. My search for a general definition of a mainframe quickly showed that finding a good definition was not going to be easy. To be sure, there were many different definitions, but they all seemed to fall into two distinct categories: broad general definitions and definitions that concentrated on some particular aspect of mainframe computing. Neither of these categories, unfortunately, provided a very good definition.

The general definitions of a mainframe most often describe only the size of the computer. For example, a mainframe is sometimes defined as a room or more of computer equipment. This definition also fits a PC server farm, and few would argue that a PC server farm is a mainframe.

Another widely used general definition describes a mainframe as a large computer for the commercial applications of Fortune 1000 businesses and other large-scale computing purposes. This definition also suffers from too many ambiguities. By this definition, any large computer that is used by a Fortune 1000 business to run commercial applications is a mainframe, and we know that is not true.

The more specific definitions of mainframe usually get too specific. They usually define a mainframe by the particular operating systems it runs. If it runs zOS, MVS/ESA (OS/390), VM/ESA, or VSE/ESA, it is a mainframe. Some of these definitions go so far as to define a mainframe as the operating system itself. Linux running on a mainframe totally confuses these definitions. If Linux is a mainframe operating system, then any computer that runs Linux must be a mainframe. Again, not true.

It should be fairly clear by now that there probably is no good definition for a mainframe that satisfies everyone. A few years back, there was a joke circulating through the various S/390 user groups around the world that has become a personal favorite of mine for the definition of a mainframe: If you can pick it up, it's a PC. If you can't pick it up, but you can push it over, it's a midrange server. But when you can't pick it up or knock it over, it's a mainframe.

At 6 feet 8 inches tall and weighing almost a ton, you are definitely not going to push over an i890. By this definition, it is a mainframe. But since few others in IBM thought this was a good definition, we decided not to call the i890 a mainframe. We also never published my whitepaper.

Finding a good definition for the iSeries entails all of the same problems as finding a good definition for a mainframe. We could simply say that an iSeries is any server that runs OS/400, but is there more to it than that?

It's Not the Hardware

One way to determine what an iSeries is is to look at what it is not. For example, it is not a hardware box. Oh, IBM does sell you a black box, and on the front of that box, it does say "IBM eServer iSeries," but the hardware is just the visible part of the iSeries. The iSeries itself is the virtual machine that resides inside the black box.

The iSeries has never been defined by its hardware. It is defined by software. This is in stark contrast to every other commercially available server, where the hardware architecture is the defining element of the server. This means that all software in these servers, including applications, is dependent on the underlying hardware architecture.

The power of not defining the server by its hardware was dramatically illustrated when the AS/400 processor technology moved from CISC to RISC in the middle 1990s. Customers saved the programs off their CISC machines, restored them on their new RISC machines, and the programs ran as full 64-bit applications without even having to be recompiled. No other server has ever made such a major hardware transition. Those that have tried failed. For example, Digital in the 1990s tried to move its VAX installed customers to RISC hardware but was not successful. Today others, including HP, are just beginning to make the transition to a new hardware base, and their success is by no means assured.

We used to talk about this design concept of a virtual machine as technology independence. In recent years, the term "virtualization" is being used more frequently to describe the use of software rather than hardware to define the interfaces to various parts of the server.

The point is that the iSeries is not defined by its hardware. That hardware can in the future change completely, and we will still have an iSeries. If it is not the hardware that defines the iSeries, it must be the software.

It's Not the Operating System

A few years back, there was a great deal of discussion in the industry about operating systems not being very important. Even though not everyone agreed, there were those who predicted that all proprietary operating systems would eventually go away. That hasn't happened yet, but there are still some people who believe it will.

Defining a server in terms of its operating system would seem to be a short-lived definition, especially if you believe that all operating systems are similar and that it is unimportant which one you choose. Without arguing for or against either side of this discussion, it is true that all operating systems do pretty much the same things.

Any computer science student can probably tell you that an operating system fundamentally does only two things. It controls all of the computer's resources, and it provides the base upon which applications are written. That student can also probably tell you that an operating system has four major components: process management, memory management, input/output, and the file system. On the surface, it seems hard to believe that one server can be distinguished from another solely on the basis of its operating system. This gives credence to the argument that one operating system is as good as another.

Notice also that the components of an operating system deal primarily with the underlying hardware. An operating system must know the details of the hardware. In other words, an operating system cannot be technology independent. Moving an operating system from one hardware platform to another requires a great deal of work, assuming you want to have an efficient implementation of that operating system. For example, it took nearly five years and hundreds of system programmers to rewrite the operating system when the AS/400 moved from CISC to RISC.

Because the operating system is dependent on the underlying hardware and must be heavily modified to run on different hardware, the iSeries is not defined by its operating system. If we were to change the operating system of the iSeries completely, it would still be an iSeries.

A Poor Choice for a Name

Now here's where it gets tricky. What is the operating system of the iSeries? If you said OS/400, you would be wrong. None of those four major components of an operating system described in the previous section are in OS/400. Simply stated, OS/400 is not an operating system, and it never has been.

Twenty-five years ago in October 1978, the IBM System/38 was announced. That system was the beginning of what has become today's iSeries. Because the S/38 was technology independent (defined by its software and not by its hardware), any component of that system that was hardware dependent was packaged as a part of the hardware. As a result, the operating system of the S/38 was packaged as part of the microcode. Microcode in those days was considered to be hardware.

Other system software that was not hardware dependent was packaged separately. This software package included the world's first commercially available relational database, functions to manage work in the system, a new control language, data definition interfaces, extensive security and authorization facilities, and some powerful program development tools. Because many of the functions in this system software package had to do with the overall control of the S/38, we decided to call the package Control Program Facility (CPF). CPF was not an operating system in the textbook sense, although it was sometimes called an operating system for lack of a better, more descriptive name. The real operating system of the S/38 was in the microcode.

In 1988, the S/38 was renamed the AS/400 yes, we have renamed this system more than once. Someone, obviously a marketing type, also decided to rename CPF to OS/400. In hindsight, this was a huge mistake. Suddenly, CPF, with all its sophisticated functions and capabilities, was degraded to the level of a simple operating system. To this day, people still try to compare OS/400 to a Unix or Windows operating system. That's like comparing an aircraft carrier to a rowboat. They both float, but similarities end there.

When we again renamed this system from AS/400 to iSeries, some of us wanted to change the name of OS/400 to show that it was far more than just a simple operating system. That name change was never made because the OS/400 name was the only link that customers, many of whom weren't too happy about the name change anyway, had to the AS/400.

Still another change made in 1988 for the AS/400 was to drop the name microcode and instead use the name licensed internal code (LIC) to describe the system software that was dependent on the hardware. Microcode is part of the hardware, and customers can own the hardware. System software, including operating systems, is generally treated as licensed software. The name change meant that the AS/400 operating system, which had previously been considered hardware, became licensed software. With the introduction of the RISC systems, that licensed software became System Licensed Internal Code (SLIC). SLIC is the operating system of the iSeries.

OS/400: The Defining Entity

There you have it. The one thing that makes an iSeries an iSeries is OS/400. The specific hardware isn't important. It has and will continue to change in the future as more and more common components are shared across IBM's servers. SLIC, the operating system, is also not the big differentiator. It has undergone major changes over time and will change even more in the future as system software components are also shared among IBM servers.

OS/400, on the other hand, is unique to the iSeries. It is OS/400 that provides the control environment and the application environment that make the iSeries something very special. It is not the simple operating system that its name implies. Maybe someday in the future we will have the opportunity to give OS/400 a name that properly reflects its true value.

Meanwhile, the next time someone tries to tell you that in the future you will be able to buy an iSeries without OS/400, set the record straight. Without OS/400, it is not an iSeries. It's just another server.

Frank G. Soltis of IBM Rochester created the technology-independent architecture used in the AS/400 and iSeries. He is IBM's iSeries chief scientist and a professor of computer engineering at the University of Minnesota.

Penton Technology Media
Connected Home | Group Computing | IT Buyer's Network | SQL Server Magazine | WebSphere Professional | Windows 2000 Magazine
Report Bugs | Contact Us | Comments/Suggestions | Site Use Agreement | Privacy Policy
Copyright Penton Technology Media
Penton Media
iSeries is a trademark of International Business Machines Corporation and is used by Penton Media, Inc., under license. is published independently of International Business Machines Corporation, which is not responsible in any way for the content. Penton Media, Inc., is solely responsible for the editorial content and control of the iSeries Network.