Time-sharing System Research Articles

Analysis of RISC performance in real‐time system

AbstractThe author has already proposed a method of analyzing the average instruction execution time for estimating the performance of CISC in a real‐time system (RTS) [1]. However, in [1] there was no mention of a method of analyzing the execution time when the CPU has a cache memory and an RISC architecture.This paper presents a simulation method for solving that problem, and clarifies the relative processing power of the two architectures. The new method converts actual RTS CISC trace data in an RTS to RISC access state data and uses the data for input in simulations of cache operation, etc., to clarify the processing performance ratio between CISC and RISC in an RTS.The results clarify that the 1‐level write‐through cache strategy (MIPS R3000) suffers from continuous writing, the two‐level write‐back cache strategy (MIPS R4000) depends strongly on the speed of the secondary cache, and that the overall RTS characteristics differ from the characteristics of application programs in a time‐sharing system (TSS), etc.These techniques go beyond simply allowing the comparison of RISC and CISC performance. They also make it possible to clarify the factors of the performance characteristics of systems that include caches on the basis of an overall software model for actual RTS.

Properties of delay-cost scheduling in time-sharing systems

We consider properties of time-sharing schedulers with operations based on an economic measure termed the delay cost, and relate these to scheduling policies such as those used in VM and MVS. One of these policies, deadline scheduling, is shown to be potentially unstable. We develop delay-cost schedulers that meet similar performance objectives under quasi-equilibrium conditions but which are stable under rapidly varying loads.

'Prestige luster' and 'snow-balling effects': IBM's development of computer time-sharing

In the middle 1960s IBM responded to pressure from its most prestigious customers to hasten the development and availability of computer time-sharing systems. When MIT and Bell Laboratories chose General Electric computers for their new time-sharing system, IBM management feared that the prestige luster of these customers would lead other customers to demand the same capabilities and that there would be a snow-balling effect as more customers rejected IBM computers. IBM worked on a time-sharing product and brought it to market by the end of the decade despite greater-than-expected costs. Meanwhile MIT, Bell Laboratories, and GE worked together on a new time-sharing system known as Multics. By examining IBM's role in and response to the development of time-sharing, the article illustrates the nontechnological criteria that even high-technology companies use to decide what products to develop and market. >

An error bound for a continuous time approximation of a time-sharing queue

Performance characteristics of systems with discrete time are often approximated by similar characteristics of corresponding (in some sense) continuous time systems. Both from practical and theoretical points of view it is very important to estimate an error of such an approximation. In a recent paper N.M.van Dijk and P.G.Taylor investigated this problem for the classic time-sharing system and obtained an error bound which is of the same order as the length of time slot. In the present paper we suggest another approach to the problem based on stochastic orderings. This approach can be applied to a wider class of performance characteristics and our bound seems to be essentially better

A LOTOS extension for the performance analysis of distributed systems

Performance analysis and formal correctness verification of computer communication protocols and distributed systems have traditionally been considered as two separate fields. However, their integration can be achieved by using formal description techniques as paradigms for the development of performance models. This paper presents a novel extension of LOTOS, one of the two formal specification languages that were standardized by ISO. The extension is specifically conceived to integrate performance analysis and formal verification. The extended language syntax and semantics are formally defined, along with a mapping from extended specifications to performance models, The mapping preserves the specified observable behavior. Two simple examples, a stop-and-wait protocol and a time-sharing system, are used to concretely demonstrate the new approach and to validate it.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Editorial: Resource allocation in Real-Time Systems

Large real-time systems are complex. In these systems there are usually many periodic and aperiodic tasks, a high degree of concurrency, and significant asynchronous behavior. Requirements often demand careful design and analysis so that the system is predictable. Unfortunately, much of the analysis has focused on a too simplistic model. For example, the basic rate monotonic result requires many simplistic assumptions which are not met in practice. Consequently, for a complex system it does not matter that you have a beautiful theorem with precise analysis because the coverage of the assumptions compared to the real system is low. While the basic approach may still be useful for an approximate analysis, it must be noted that any approximate analysis is more limited for a real-time system than for a general purpose timesharing system because designers require precise analysis with respect to deadlines of specific tasks. Another example of simplistic analysis is earliest deadline scheduling. Simple formulas can again provide precise results based on assumptions such as all tasks are preemptable at any point, there is no sharing of resources, and no overloads will ever exist. Again, if the manner in which these assumptions match the real system is poor then the precise analysis is not precise at all for the system in question! One key problem with the basic rate monotonic and earliest deadline policies and their analyses is that they fail to recognize that all resources must be addressed by the analysis and not just the CPU. In other words, what is required is an integrated resource allocation model for real-time where

Impact of the Personal Computer on X-Ray Analysis Historical Perspective 1960-1990

In these modern times, where the use of the computer in the analytical laboratory is taken for granted, it is perhaps difficult to realize that, less than one generation ago, computers were little more than an idea on an engineer's desk. It is interesting to note the sequence in which the automation of data collection and data processing developed. As would be expected, the time sequence followed closely the developments in computer hardware and peripherals. An important factor in the development of most commercial automated systems was the “20%” rule. This rule required that the total cost of any computer package should not exceed 20% of the sale price of the final automated product. Rex's “Numerical Control Powder Diffractometer” was described in the 1966 Denver Conference and this machine was to be the forerunner of a whole host of automated diffractometers which appeared in the early 1970s. Typical systems used either a 4K minicomputer or a time-sharing system with a large main-frame computer. It is interesting to observe that, as we come into the 1990s, the argument as to whether the main-frame will survive as a viable alternative to the rapidly developing PC still goes on.

Research guide to corporate acquisitions, mergers, and other restructuring

Preface Coping with Corporate Change Monitoring Corporate Change with Print Resources Monitoring Corporate Change with Electronic Resources Williams Act Filings: Public Disclosure of M&A M&A Transaction Databases Finding Acquisition Candidates The Muddled Corporate Identity International M&A: A Cross-Border Perspective Information Checklist Appendix 1: Online Database Chart Appendix 2: Directory of Databases Time-Sharing Systems Glossary Select Bibliography Index

The Rise and Fall of a PM System

ABSTRACTThe Sundance Thermal Generating Plant has six 375-megawatt generators. Before 1979, its annual maintenance shutdowns were handled entirely by external specialized contractors. In 1980, the plant staff started to assume project management control of the shutdowns, using the contractors as subcontractors. Working initially with a training consultant, the plant crew started with short planning sessions and manually updated CPM networks. Over the next 4 years, the project management methods quickly advanced to a simple computer critical path on a time-share system, and then to a complex resource-balanced program on a dedicated VAX. By this time, most of the contractors had been replaced by locally hired tradespeople who were managed directly by the plant staff.The consultant left the project in 1984, and the maintenance staff took over the training and development tasks associated with the system. Seven years later, in 1992, the project management system was observed, evaluated, and compared against t...

Computerized Prescription Writing in Otolaryngology

To the Editor .—Pioneering efforts to computerize prescription writing have been hindered by hardware limits. Most hospitals and institutions used time-sharing systems, and the first computers were expensive, inflexible, and cumbersome. Personal computers (PCs) have developed greatly during recent years as regards memory, speed, and computing ability, minimizing the above problems. Taking advantage of this progress we have designed a PC-based prescription-writing program for use by otolaryngologists. The program is currently running in an IBM PC with a 20-megabyte fixed disk, 640 kilobytes of random access memory (RAM), and a video graphics adaptor monitor. However, the program will operate on any IBM-compatible PC, needing only 64 kilobytes of RAM. Prescriptions are printed on continuous fanfolded narrow paper by a dot-matrix printer. The program was written using random files of BASIC, the most friendly and widespread programming language. Three files have been created: the drug database, the main program, and a

Prolog to the future

The series of 'Time-Sharing System Scorecards' published from 1965-7 by the Computer Research Corp. chronicles the development of time-sharing systems in research organizations (universities and laboratories) and the expectations for commercial offerings. The last of these scorecards (reproduced in the article) shows a stage of development when time-sharing was still new enough to be a research effort and not quite mature enough to be commonplace. >

Project MAC (time-sharing computing project)

The history of Project MAC from its inception is reviewed. It is noted that the story of Project MAC is a continuation of that of CTSS (Compatible Time-Sharing System), marking the transfer of an experimental system into the practical academic field and thence into the commercial field. Two men were key to this recognition of the value of the work by Corbato et al.-J.C.R. Licklider and R.M. Fano. Licklider's early concepts of man-computer symbiosis were identified as being supportive of the ideas which led to the development of CTSS, and his relocation to the Advanced Research Projects Agency (ARPA) puts him in a position to do something about his aspirations. Excerpts from reports dealing with Project MAC are presented. >

CTSS-the compatible time-sharing system

Excerpts are presented from a 1962 paper by Fernando Corbato, M. Merwin-Daggett, and R.C. Daley. The selection from the paper, which reported work completed in late 1961, is reprinted from Proc. Spring Joint Computer Conf., vol.21, 1962. In this system, the time-sharing occurs among four users, three of whom are online each at a typewriter in a foreground system, and a fourth passive user of the background FAP-MAD-MADTRAN-BSS Monitor System used by most of the MIT Computation Center programmers and by many other 7090 installations. >

Machine Tongues XIII: Real-Time Audio Conversion under a Time-Sharing Operating System

Time-critical processing of data is a necessary part of many applications written for UNIX and other time-sharing system environments. Although it is known that standard UNIX is not a real-time operating system in any sense, applications that run under UNIX may achieve a measure of time-critical performance by exploiting or bypassing certain aspects of the operating system. Digital-to-analog conversion of sound data is a process by which a numerical representation of a sound wave is translated into the corresponding sound signal by specialized hardware. It is an example of an application in which hardware and software must cooperate closely to meet hard deadlines for data availability. A steady stream of sound data samples must reach the sound-conversion hardware from some storage medium interfaced to the host system, and this must occur at a rate sufficient to sustain playback of the sound signal at the desired rate. Failure to meet these requirements can result in a sound stream full of clicks, skips, or silences. Digital representation and processing of sound requires considerable space for data as well as system speed. According to the Nyquist theorem, a complex sound wave must be sampled at-at least-twice its greatest frequency, in order to be accurately reproduced. Quality recording and playback of sound generally means a sample rate on the order of 48 kHz, each sample having at least 16 bits of precision. Stereo sound requires twice as many samples as mono. So, in these terms, a stereo recording 5 minutes in duration would occupy almost 60 MBytes of disk storage. A 20-minute piece of music would require about 236 MBytes of disk storage. Delivery of these data to the convertor hardware at a rate necessary to maintain continuous playback of sound can depend on several factors, including the speed of access to sound data stored on a disk, the effective bandwidth of the bus on which data are delivered to the convertors, the hardware design of the device that controls the transfer of data from host memory to convertors, and the software design of its device driver. If the convertor hardware is able to store 48,000 samples (i.e., 96,000 bytes) of sound data, for example, 250 msec of stereo sound can be buffered at the 48-kHz sampling rate. Once audio playback begins, there would be less than 250 msec in which to make the next buffer of sound data available to the convertors before the current buffer empties. This is an example of a hard deadline. A hypothetical bus with disk subsystem attached might have the ability to transfer data to the convertors at a rate of 1 MByte/sec, a conservative bandwidth by today's standards. Transferring 48,000 bytes of data in 250 msec would be well within the abilities of such an I/O path as judged by transfer rate specifications alone. However, in a multiuser, multitasking environment, concurrent system activities take bus bandwidth, disk cycles, and processor time away from the audio playback process, interfering with sustained throughput. It is throughput, rather than maximum burst rate of an I/O path, which determines whether playback of high-quality stereo sound will be continuous or broken. It takes only one late sample to produce a gap in the sound as it is heard, no matter how many previous samples might have arrived on time. At the Center for Music Experiment (CME) of the University of California, San Diego, I have had the opportunity to work with three commercially available sound-conversion systems running under the UNIX operating system. All three systems claim to deliver 16-bit, 48-kHz stereo playback of sound in a regular multiuser UNIX environment. One system Computer Music Journal, Vol. 15, No. 3, Fall 1991, ? 1991 Massachusetts Institute of Technology.

A retrospective on the VAX VMM security kernel

The development of a virtual-machine monitor (VMM) security kernel for the VAX architecture is described. The focus is on how the system's hardware, microcode, and software are aimed at meeting A1-level security requirements while maintaining the standard interfaces and applications of the VMS and ULTRIX-32 operating systems. The VAX security kernel supports multiple concurrent virtual machines on a single VAX system, providing isolation and controlled sharing of sensitive data. Rigorous engineering standards were applied during development to comply with the assurance requirements for verification and configuration management. The VAX security kernel has been developed with a heavy emphasis on performance and system management tools. The kernel performs sufficiently well that much of its development was carried out in virtual machines running on the kernel itself, rather than in a conventional time-sharing system. >

Experiences with the Amoeba distributed operating system

The Amoeba project is a research effort aimed at understanding how to connect multiple computers in a seamless way [16, 17, 26, 27, 31]. The basic idea is to provide the users with the illusion of a single powerful timesharing system, when, in fact, the system is implemented on a collection of machines, potentially distributed among several countries. This research has led to the design and implementation of the Amoeba distributed operating system, which is being used as a prototype and vehicle for further research. In this article we will describe the current state of the system (Amoeba 4.0), and show some of the lessons we have learned designing and using it over the past eight years. We will also discuss how this experience has influenced our plans for the next version, Amoeba 5.0. Amoeba was originally designed and implemented at the Vrije Universiteit in Amsterdam, and is now being jointly developed there and at the Centrum voor Wiskunde en Informatica, also in Amsterdam. The chief goal of this work is to build a distributed system that is transparent to the users. This concept can best be illustrated by contrasting it with a network operating system, in which each machine retains its own identity. With a network operating system, each user logs into one specific machine—his home machine. When a program is started, it executes on the home machine, unless the user gives an explicit command to run it elsewhere. Similarly, files are local unless a remote file system is explicitly mounted or files are explicitly copied. In short, the user is clearly aware that multiple independent computers exist, and must deal with them explicitly. In contrast, users effectively log into a transparent distributed system as a whole, rather than to any specific machine. When a program is run, the system—not the user—decides upon the best place to run it. The user is not even aware of this choice. Finally, there is a single, system-wide file system. The files in a single directory may be located on different machines, possibly in different countries. There is no concept of file transfer, uploading or downloading from servers, or mounting remote file systems. A file's position in the directory hierarchy has no relation to its location. The remainder of this article will describe Amoeba and the lessons we have learned from building it. In the next section, we will give a technical overview of Amoeba as it currently stands. Since Amoeba uses the client-server model, we will then describe some of the more important servers that have been implemented so far. This is followed by a description of how wide-area networks are handled. Then we will discuss a number of applications that run on Amoeba. Measurements have shown Amoeba to be fast, so we will present some of our data. After that, we will discuss the successes and failures we have encountered, so that others may profit from those ideas that have worked out well and avoid those that have not. Finally we conclude with a very brief comparison between Amoeba and other systems. Before describing the software, however, it is worth saying something about the system architecture on which Amoeba runs.

Queues with Slowly Varying Arrival and Service Processes

We examine a generalisation of the M/G/1 queue. The arrival and service processes are governed by a Markov chain which determines the rate of arrival and the service time distribution from a finite set. This Markov chain is assumed to vary “slowly”, so that we are able to derive analytical results for the stationary distribution of the queue length using an approach based on decomposability. The practical interest of this model stems from the numerous applications where the parameters of queueing systems are time varying, such as inventory models, telephone systems, time-sharing systems, computer networks with bursty traffic, etc. We also show how this approach can be extended to arbitrary networks of queues and in particular to those with product form solution.

An introductory psychology laboratory system featuring inexpensive time-shared record-keeping

At California State University, Fresno, the introductory psychology lecture sections operate independently of the laboratory system. Each week, the faculty supervisor prepares the teaching assistants to instruct the following week’s laboratory presentation. The department’s subject pool is also administered through the laboratory system. At the end of the semester, the lecture instructors are supplied with laboratory grades for incorporation into the overall course grades. Record-keeping, especially for the subject pool, was a major problem until we computerized its various components. We use one of the university’s time-sharing systems for this purpose, which has proven cost effective, secure, and quite efficient.

Analytic Models of Supercomputer Performance in Multiprogramming Environments

Supercomputers run multiprogrammed time-sharing operating systems, so their facilities can be shared by many local and remote users. Therefore, it is important to be able to assess the performance of supercomputers in multiprogrammed environments. Analytic models based on Queueing Networks (QNs) and Stochastic Petri Nets (SPNs) are used in this paper with two purposes: to evaluate the performance of supercomputers in multi programmed environments, and to compare, perfor mance-wise, conventional supercomputer architectures with a novel architecture proposed here. It is shown, with the aid of the analytic models, that the proposed architecture is preferable performance-wise over the existing conventional supercomputer architectures. A three-level workload characterization model for super computers is presented. Input data for the numerical ex amples discussed here are extracted from the well- known Los Alamos benchmark, and the results are vali dated by simulation.

Remote site treatment planning

Remote site treatment planning was one of the earliest methods of digital computer dose planning. Using batch-oriented or time-sharing mainframe computer systems, researchers in the late 1950s and early 1960s developed many of the basic algorithms used later in stand-alone systems. Although time-shared systems have continued in use, most of the emphasis on treatment planning in the last decade has been on using dedicated mini- and micro-computer systems. Recent developments in the computer industry, such as the use of networks and distributed computer processing, may lead to a resurgence of interest in remote computer systems for treatment planning.