| Previous | Table of Contents | Next |
Time is Mother Nature’s way to keep everything from happening at once. Processes are a computer’s way of doing the same thing. A process is a program in execution, which consists of the executable program, the program’s data, and any state information (defined below) needed to continue running the program. Every operating system provides the facilities to support a process. Earlier, we said a process could be thought of as a unit of work in the system, and this is still a good definition.
Perhaps the easiest way to get an intuitive feel for a process is to think about timesharing systems. Timesharing, as we saw in Chapter 8, means sharing the processor and memory with several users at the same time to give the appearance that every user has his or her own machine. With a single processor, only one user process can be executing a program at any instant in time. Process management is the SLIC component that keeps everything from happening at once by switching the resources of that single processor between processes.
Periodically, the operating system decides to stop running one process and start running another because, for example, the first one has used its current share of processor time as of the last second. When a process is temporarily suspended like this, it must be restarted later in exactly the same state it had when it was stopped. This means all information about the process — called the state information — must be saved somewhere during the suspension. How and where this state information is stored varies from one operating system to another. In this chapter, we will see how the AS/400 does it.
OS/400’s work-management component deals with these same topics at a higher level. The need to efficiently manage the flow of work through and within the system is important to achieve high performance for a wide range of application environments. Later in this chapter, we will see how work management interrelates with process management, but first we look at the foundation on which process management is based.
In any computer system, often only a small number of basic ideas give the system a competitive edge. The technology independence of the AS/400 the MI provides is one such idea, as is the performance advantage the single-level store provides. Some of these ideas are known to everyone; others are not. One of the most important ideas in the AS/400 is the tasking structure, yet it is not widely known.
IBM and other companies always want to be sure the important ideas in any new product are protected by patents, so others cannot copy them. Patents give a company exclusive rights to specific new ideas, which can give the owning company an advantage over its competitors if the new idea is significant enough. These key patents are also valuable in the respect that the rights to use the patent can be sold to other companies. Before the AS/400’s announcement, a study of the patents covering the new system was undertaken. The intent of this study was to select the most important of these patents.
The most important patent on the AS/400 was U.S. Patent Number 4,177,513,1 which covered the tasking structure of the System/38 and the AS/400. (The patents covering single-level store were not selected, nor were any patents on technology independence.)
1Roy L. Hoffman, William G. Kempke, John W. McCullough, Frank G. Soltis, and Richard T. Turner, “Task Handling Apparatus for a Computer System,” United States Patent 4,177,513, filed July 8, 1977, issued December 4, 1979.
The tasking structure of the AS/400 is the foundation on which the operating system is built. SLIC’s process-management component and OS/400’s work-management component are based on this tasking structure. For most of the operating system topics covered so far, we have started at the highest level in the system and progressively worked our way down to the lowest level of support. In this chapter, we want to start at the bottom, because it is so fundamental to the AS/400. Before we do so, a few words are in order about future directions for operating systems. Only then will the importance of the tasking structure become obvious.
A microkernel is a small operating-system core and one of the most hotly debated topics in computer science today. Proponents of microkernels claim that they provide the foundation for modular, portable operating systems. Opponents of microkernels claim they create a bottleneck in the operating system that limits the size of a multiuser system. Almost no one agrees how to organize the operating-system services relative to the microkernel. The one thing experts seem to agree on is the message-passing communications structure a microkernel uses. Most believe this is the direction for all future operating systems, whether or not they are microkernel-based.
To better understand this message-passing communications structure, let’s look briefly at how operating-system components have traditionally communicated and interacted. A Unix operating system provides us with a prime example. The original Unix, and most of the versions currently available, use a layered approach to operating-system design.
Groups of operating-system functions in Unix, such as the file subsystem, the process control subsystem, and the I/O subsystem, are divided into layers. This layering approach is not unusual. Most operating systems, including the one in the AS/400, are built using layers of software. The differences in design among different operating systems relate to the way the layers communicate and work with one another. In a Unix system, each layer communicates only with the layer directly above it or that below it. This is an advantage, because each layer needs to know about only the ones immediately above or below it. Requests and responses move from layer to layer as if they are climbing up and down a ladder. Applications and the operating system itself communicate with other functions using this ladder.
| Previous | Table of Contents | Next |