Multiprogramming System Research Articles

Multiserver Queueing Models of Multiprocessing Systems

Conventional time sharing and multiprogramming systems have been extensively modeled as single-server queues. In contrast, multiprocessing systems must be modeled as multiserver queueing systems. This paper investigates the effect of the scheduling strategy-a key parameter-on the performance of multiserver queues, under very general assumptions: arbitrary arrival process and arbitrary service times distribution. We distinguish two classes of queue disciplines: preemptive and nonpreemptive, and two types of arrival processes: dependent (on the state of the queue) and independent. The performance is evaluated in terms of global measures: the utilization factor of the servers, as a measure of internal efficiency, and the mean response time, as a measure of external (user-oriented) performance. It is shown that nonpreemptive disciplines leave both measures invariant, for given but arbitrary (dependent or not) arrival process and arbitrary service distribution. For preemptive disciplines, the utilization factor is invariant only if the arrival process is independent, while the mean response time depends on the discipline, whatever the arrival process. It is also shown that the exponential distribution is the only distribution for which the queue discipline does not affect the performance of the system, thus underlining the very peculiar nature of this standard assumption and preventing us from safely generalizing the conclusions reached under this assumption. A quantitative comparison of preemptive and nonpreemptive disciplines requires an analytical solution, for which a general technique is presented.

Oscar: A multiprogrammed small computer system for experimental nuclear physics

The central component of the data acquisition and analysis system at the Ohio State University Van de Graaff Laboratory is an IBM 1800 computer, operating under a powerful multiprogramming monitor. The system has been designed to allow efficient usage of the computer resources by several experimentalists at once. Under typical conditions, the processor is shared concurrently by data acquisition and display routines, an “on-line” data reduction or experimental control program, and an “off-line” analysis or system updating job. Operations are enhanced by two interactive display units. The system has been found to be flexible, easy to program, and straightforward to interface.

The Effect on Throughput of Multiprocessing in a Multiprogramming Environment

This paper investigates some of the effects of varying the number of central processing units (CPU's) available to a multiprogramming system both when parallel processing of a single task (multitasking) is allowed and is not allowed. The variables investigated are those that would be expected to control the CPU queue length distribution and include the degree of multiprogramming, CPU service discipline, CPU service distribution, the degree of cooperation between processors engaged in multitasking, and the ratio of CPU to I/O service capacity. The computer systems are modeled by queueing networks. Analytic results are displayed in many cases with supplementation by numeric and simulation solutions when convenient or necessary. It is demonstrated that under properly defined conditions doubling the number of CPU's in a system can more than double throughput. The effectiveness of multitasking in improving throughput is found to be fairly small if the degree of multiprogramming ranges from modest to high.

Efficient multiprogramming resource allocation and accounting

Although sometimes thought of as only a component of time-sharing operation, multiprogramming can involve broader questions of resource allocation, since fairness is not required to meet a response criterion. In a multiprogrammed system, it may serve maximal resource use to be unfair, for example by holding an input/output channel idle for a program while it completes a small amount of processor usage, enabling further use of the channel. Several applications of this principle are given, and it is suggested that a multiprogramming executive might dynamically adjust its allocation algorithms to gain efficiency. Allocation of resources is closely connected to accounting for those resources, raising the problems of repeatability, minimal uncharged overhead, and relative weighting of charges for dependent resources. Since weightings may depend on allocation algorithms, these are not arbitrary accounting parameters. Often the only repeatable accounting is one which omits an extensive overhead; this overhead can be multiple-charged to all programs which benefit from it, so that in the worst case the overhead will be paid, and should multiprogramming prove efficient, overcharges will result. Multiprogramming turns on allocation of the memory resource essential to control of other resources. The general suggestions for allocation and accounting are applied to this question, and some details provided for the case of a monitor which controls a virtual-memory machine.

Processor Utilization in Multiprogramming Systems via Diffusion Approximations

Cyclic queuing systems have been proposed by several authors in the study of the behavior of multiprogrammed computer systems. Programs in the system wait for service at the central processor unit (CPU); then, after page fault or input-output request at a data transmission unit (DTU), the process repeats until the program completes. Semi-Markov analysis of such systems, based on the apparently plausible assumption of independently but exponentially distributed CPU burst time, and independent, but nearly constant DTU tune may be conducted. This paper presents some very simple approximations based on a continuous-state approximation—the simple diffusion with two reflecting barriers—to describe the CPU utilization. Computational experience from which the quality of the approximations can be assessed is reported.

The people's time sharing system

AbstractA set of programs running under a multiprogramming batch operating system on the CDC 6600 which provide remote users with a time sharing service is described. The basis for the system is the ability of a user program to create job control statements during execution, thereby tricking the operating system into treating it as an ordinary batch job. The text editor and the interactive debugging facilities are described. The performance of the system, known as the People's Time Sharing System (PTSS), and user reaction to it are also described.

An on-line multi-instrument laboratory computer system

An on-line computer system is described that concurrently performs data acquisition from several laboratory instruments and experiments. The computer, a 16 bits/word minicomputer, is equipped with a multiprogramming operating system and operates on a 24 h/day basis. Hardware and software elements of the system, related to the research laboratory requirements, are described. Advantages and disadvantages of the multiprogramming approach with extensive use of FORTRAN are presented and some of the on-going biochemical and physiological applications discussed.

Testing a multiprogramming system

AbstractA central problem in program design is to structure a large program such that it can be tested systematically by the simplest possible techniques. This paper describes the method used to test the RC 4000 multiprogramming system. During testing, the system records all transitions of processes and messages between various queues. The test mechanism consists of fifty machine instructions centralized in two procedures. By using this mechanism in a series of carefully selected test cases, the system was made virtually error free within a few weeks. The test procedure is illustrated by examples.

A task-scheduling algorithm for a multiprogramming computer system

This paper presents a description and analysis of a task scheduling algorithm which is applicable to third generation computer systems. The analysis is carried out using a model of a computer system having several identical task processors and a fixed amount of memory. The algorithm schedules tasks having different processor-time and memory requirements. The goal of the algorithm is to produce a task schedule which is near optimal in terms of the time required to process all of the tasks. An upper bound on the length of this schedule is the result of deterministic analysis of the algorithm. Computer simulations demonstrate the applicability of the algorithm in actual systems, even when some of the basic assumptions are violated.

Microwave data link for computer communication

An experiment is described in linking to a large digital computer at data rates in the order of 107 bit/s. The system is based on standard British Post Office microwave equipment and well established high-speed wire links; the interface between the two sections uses the CAMAC-system convention. The total system is described, and details are given of some of the equipment not previously published. Important features of the system are the termination of both ends of the link in a high-level-language module (in this case PL1), executing in a multiprogramming operating system, and corrected error rates better than 1 bit in 109.

Dynamic storage partitioning

A model of paged multiprogramming computer systems using variable storage partitioning is considered. A variable storage partitioning policy is one which allocates storage among the active tasks according to a sequence of fixed partitions of main storage. The basic result obtained is, mean processing efficiency is increased and mean fault-rate decreased under a variable partition, provided that the curves of efficiency and fault-rate as a function of allocated space are concave up.

A queuing model of a multiprogrammed computer with a two-level storage system

The results are presented of an analysis of a probabilistic model of a multiprogrammed computer system with a two-level storage system in which there is sequential dependency of accesses between the devices. Expressions are obtained for the long-run probability that both the CPU and each of the storage devices are busy. Some numerical results are given which quantify the gains in CPU utilization obtainable by multiprogramming in the presence of this type of storage system.

A page allocation strategy for multiprogramming systems

In a multiprogramming, virtual-memory computing system, many processes compete simultaneously for system resources, which include CPU's, main memory page frames, and the transmission capacity of the paging drum. (We define a “process” here as a program with its own virtual memory, requiring an allocation of real memory and a CPU in order to execute). This paper studies ways of allocating resources to processes in order to maximize throughput in systems which are not CPU-bound. As is customary, we define the multiprogramming “set” (MPS) as the set of processes eligible for allocation of resources at any given time. Each process in the MPS is allocated a certain number of page frames and allowed to execute, interrupted periodically by page faults. A process remains in the MPS until it finishes or exhausts its “time slice”, at which time it is demoted. We assume the existence of two resource managers within the operating system: The Paging Manager and the Scheduler. The function of the Paging Manager is to control the size of the MPS, and to allocate main storage page frames among those processes in the MPS. The function of the Scheduler is to assign time-slice lengths to the various processes, and to define a promotion order among those processes not currently in the MPS. The Scheduler must ensure that system responsiveness is adequate, while the Paging Manager is primarily concerned with throughput. This paper studies possible strategies for the Paging Manager. A strategy for the Scheduler is proposed in (2).

Trace-driven modeling and analysis of CPU scheduling in a multiprogramming system

Microscopic level job stream data obtained in a production environment by an event-driven software probe is used to drive a model of a multiprogramming computer system. The CPU scheduling algorithm of the model is systematically varied. This technique, called trace-driven modeling, provides an accurate replica of a production environment for the testing of variations in the system. At the same time alterations in scheduling methods can be easily carried out in a controlled way with cause and effects relationships being isolated. The scheduling methods tested included the best possible and worst possible methods, the traditional methods of multiprogramming theory, round-robin, first-come-first-served, etc., and dynamic predictors. The relative and absolute performances of these scheduling methods are given. It is concluded that a successful CPU scheduling method must be preemptive and must prevent a given job from holding the CPU for too long a period.

Locality in Page Reference Strings

A probabilistic model is presented of program material in a paging machine. The sequences of page references in the model are associated with certain sequences of LRU stack distances and have reference patterns formalizing a notion of “locality” of reference. Values for parameters of the model can be chosen to make the page-exception characteristics of the generated sequences of page references consistent with those of actual program traces.The statistical properties of the execution intervals (times between page-exception) for sequences of references in the model are derived, and an application of these results is made to a queuing analysis of a simple multiprogrammed paging system. Some numerical results pertaining to the program model and the queuing analysis are given.

On Memory Usage Under Microprogramming

The issue of quantitative determination of memory usage and utilization is central to multiprogramming computer systems. In such systems, one tries to improve the system performance by loading a set of independent jobs together into main memory. If the jobs have complementary resource requirements and the system is capable of parallel resource operation, the total time required to complete all jobs is less than if they were processed serially, i.e., in a uniprogramming mode. The situation is illustrated in Figure 1 showing a space-time diagram of memory occupancy, where the available memory space is dynamically covered with jobs. The term job is used in the sense of process as defined in [1], which also gives an excellent survey of related subjects. However, we restrict jobs to keep a fixed amount of memory until they are completed. Generalization of the concepts to be developed to include systems with dynamic allocation, swapping, paging is possible but not attempted here.

A model of memory contention in a paging machine

This paper is concerned with certain aspects of contention for main memory resources in a multiprogrammed computer system operating under demand paging. In the model presented, the number of page-frames of main memory allocated to a problem program varies in time. These changes in memory configuration are represented explicitly in the model, CPU requirements and page exception characteristics of program material being described statistically. Expressions for the distribution of the number of page-frames allocated to an executing program, the long run expected fraction of a program's execution time in a given number of page-frames, and the average execution interval of the multi-programmed load are obtained. It is pointed out heuristically and demonstrated numerically that an increase is obtainable in the average execution interval of the multiprogrammed load over that resulting from equal fixed partitioning of main memory.

Programming systems and languages 1965-1975

In spite of impressive gains by PL/I, Fortran and Cobol remain the languages in which most of the world's production programs are written and will remain so into the foreseeable future. There is a great deal of theoretical interest in Algol 68 and in extensible languages, but so far at least they have had little practical impact. Problem-oriented languages may very well become the most important language development area in the next five to ten years. In the operating system area all major computer manufacturers set out to produce very ambitious multiprogramming systems, and they all ran into similar problems. A number of university projects, though not directly comparable to those of the manufacturers, have contributed greatly to a better understanding of operating system principles. Important trends include the increased interest in the development of system measurement and evaluation techniques, and increased use of microprogramming for some programming system functions.

Nonpriority Multiprogramming Systems Under Heavy Demand Conditions—Customers' Viewpoint

Nonpriority Multiprogramming Systems Under Heavy Demand Conditions—Customers' Viewpoint

Short-term scheduling in multiprogramming systems

This paper defines a set of scheduling primitives which have evolved from multiprogramming systems described by Dijkstra, Lampson, Saltzer, and the present author. Compared to earlier papers on the same subject, the present one illustrates a more concise description of operating system principles by means of algorithms. This is achieved by (1) describing the primitives as the instructions of an abstract machine which in turn is defined by its instruction execution algorithm; (2) introducing a notation which distinguishes between the use of synchronizing variables (semaphores) to achieve mutual exclusion of critical sections, and to exchange signals between processes which have explicit input/output relationships; (3) considering the influence of critical sections on preemption and resumption; and (4) using a programming language, Pascal, which includes natural data types (records, classes, and pointers) for the representation of process descriptions and scheduling queues. The algorithms are written at a level of detail which clarifies the fundamental problems of process scheduling and suggests efficient methods of implementation.