The supported methodology evolving in the SARA (System ARchitects' Apprentice) system creates a design framework on which increasingly powerful analytical tools are to be grafted. Control flow analyses and program...
详细信息
ISBN:
(纸本)9781450374880
The supported methodology evolving in the SARA (System ARchitects' Apprentice) system creates a design framework on which increasingly powerful analytical tools are to be grafted. Control flow analyses and program verification tools have shown promise. However, in the realm of the complex systems which interest us there is a great deal of research and development to be done before we can count on the use of such powerful tools. We must always be prepared to resort to experiments for evaluation of proposed designs. This paper describes a fundamental SARA tool, the graph model simulator. During top-down refinement of a design, the simulator is used to test consistency between the levels of abstraction. During composition, known building blocks are linked together and the composite graph model is tested relative to the lowest top-down model. Design of test environments is integrated with the multilevel design process. The SARA methodology is exemplified through design of a higher level building block to do a simple FFT. � acm. All rights reserved.
This paper describes a recent experience in benchmarking, tuning and modelling Digital Equipment Corporation's VMS executive running on their VAX-II/780 computer. Although we emphasize modelling here, the three as...
详细信息
ISBN:
(纸本)9781450374880
This paper describes a recent experience in benchmarking, tuning and modelling Digital Equipment Corporation's VMS executive running on their VAX-II/780 computer. Although we emphasize modelling here, the three aspects are closely interrelated. The benchmarking and tuning efforts are of interest because: - They were conducted without special tools, using only the script driver and measurement packages delivered with the system. - They revealed several significant aspects of system behavior that in all probability would have been abstracted away in the course of the modelling effort, had it not been preceded by detailed benchmarking and tuning. - A four-fold average improvement in benchmark performance was achieved, with greater improvements for specific customer classes. The modelling effort is of interest because: - Conventional hierarchical modelling techniques are not directly applicable to a system in which a single disk is used for paging, swapping, and user I/O activity. In particular, swapping is significant both at a high level, where it delays a customer's admission to the multiprogramming set, and at a low level, where it affects the throughput of paging and user I/O requests. - The multiplicity of memory management mechanisms in VMS lead to behavior that is difficult to comprehend, further increasing the challenge in calculating load-dependent throughput rates. - The resulting model demonstrates good agreement with measured performance over a fairly wide range of conditions. � acm. All rights reserved.
The proceedings contain 32 papers. The topics discussed include: some experiments in directory organization - a simulation study;file allocation in distributed systems;on the self contained modeling of db/dc systems;p...
The proceedings contain 32 papers. The topics discussed include: some experiments in directory organization - a simulation study;file allocation in distributed systems;on the self contained modeling of db/dc systems;probabilistic models of inverted file information retrieval systems;multitype multiprogramming: probability models and numerical procedures;validation of a computer performance model of the exponential queuing network family;an analytic performance model of a multi-programmed batch-timeshared computer;and measurements of major locality phases in symbolic reference strings.
We develop a method based on diffusion approximations in order to compute, under some general conditions, the queue length distribution for a queue in a network. Applications to computer networks and to time-sharing s...
ISBN:
(纸本)9781450374972
We develop a method based on diffusion approximations in order to compute, under some general conditions, the queue length distribution for a queue in a network. Applications to computer networks and to time-sharing systems are presented.
In this paper, we study several probabilistic models of inverted file information retrieval systems. In particular, the portion of the system which performs the tasks of list accessing and merging is modeled. The aver...
详细信息
ISBN:
(纸本)9781450374972
In this paper, we study several probabilistic models of inverted file information retrieval systems. In particular, the portion of the system which performs the tasks of list accessing and merging is modeled. The average response times of the system for three different methods of managing the disk access and merge processor queues are discussed.
A number of laws are derived which establish relationships between throughput, response time, device utilization, space-time products and various other factors related to computer system performance. These laws are ob...
详细信息
ISBN:
(纸本)9781450374972
A number of laws are derived which establish relationships between throughput, response time, device utilization, space-time products and various other factors related to computer system performance. These laws are obtained by using the operational method of computer system analysis. The operational method, which differs significantly from the conventional stochastic modeling approach, is based on a set of concepts that correspond naturally and directly to obser-ved properties of real computersystems. Except for measurement errors, the operational laws presented in this paper apply with complete precision to all collections of observational data, and they are similar to fundamental laws found in other areas of engineering and applied science.
The paper presents an analytic performance model of a multi-programmed batch and timeshared context swapped system. The model predicts resource utilizations, batch throughput and timesharing response times as function...
详细信息
ISBN:
(纸本)9781450374972
The paper presents an analytic performance model of a multi-programmed batch and timeshared context swapped system. The model predicts resource utilizations, batch throughput and timesharing response times as functions of the system hardware characteristics, including core size;load characteristics, differing between batch and interactive users;and the numbers of batch and interactive users. Thus, it allows one to study the effects of different hardware configurations and/or load patterns on system performance.
Emulation of systems makes it possible to combine the predictive power of simulation with the advantages of measurement carried under a real system workload. An emulator is a micro-programmed implementation of the bas...
ISBN:
(纸本)9781450374972
Emulation of systems makes it possible to combine the predictive power of simulation with the advantages of measurement carried under a real system workload. An emulator is a micro-programmed implementation of the basic hardware machine. It can be easily instrumented to collect performance statistics on the instruction set processor (ISP) level and support performance measurement of different configurations and software of the emulated system. This paper describes the monitoring capabilities of the Micro-programmable Multi-Processor (MMP), a powerful emulator system that serves as an experimental tool for evaluating computersystems. The measurement capabilities of the MMP on various system levels are described, as well as existing performance monitoring tools and their applications. Preliminary results contrasting the Gibson mix and measured instruction frequencies on the AN/GYK-12 computer in a TACFIRE system are given.
This paper reports on a measurement study of the Scheduler in the PDP-]0 TENEX Operating System under normal loads in a computer Science research environment at Case Western Reserve University. A complete description ...
详细信息
This paper reports on a measurement study of the Scheduler in the PDP-]0 TENEX Operating System under normal loads in a computer Science research environment at Case Western Reserve University. A complete description of the methodology and the results of a Covariance Analysis is presented here. The analysis was made in parallel with the analysis of the collected data, therefore a brief description of the experiments as well as some comments of the most interesting results are also discussed in this report. The complete report of the study is in Gonzalez [5] which shall serve as a reference document for this paper.
A relatively high level analytical model for computersystems serving both batch and interactive users is presented. The model is unusual in its employment of an endogenous priority scheme to represent a class of stra...
详细信息
ISBN:
(纸本)9781450374972
A relatively high level analytical model for computersystems serving both batch and interactive users is presented. The model is unusual in its employment of an endogenous priority scheme to represent a class of strategies for controlling service to the two types of customers. Numerical methods developed by V. L. Wallace are used to generate steady state probability distributions for the infinite state Markov chain formed by the model. Data from the Michigan Terminal System, which includes a load controlling mechanism of the type modelled, is used to validate the model. Finally, additional parameter studies indicate that the model reflects the dynamic behavior of such system in a reasonable way.
暂无评论