Showing papers on "Server published in 1984"

Optimal control of a queueing system with two heterogeneous servers

Abstract: The problem considered is that of optimally controlling a queueing system which consists of a common buffer or queue served by two servers. The arrivals to the buffer are Poisson and the servers are both exponential, but with different mean service times. It is shown that the optimal policy which minimizes the mean sojourn time of customers in the system is of threshold type. The faster server should be fed a customer from the buffer whenever it becomes available for service, but the slower server should be utilized if and only if the queue length exceeds a readily computed threshold value.

On the optimality of the shortest line discipline

Abstract: We consider a many server queueing system with identical exponential servers. Customers arrive according to an arbitrary arrival process and upon arrival the customer must be assigned to some server's queue. We show that the shortest line discipline is the best decision rule when assigning the customers to the servers in the sense that the corresponding departure process is stochastically fastest.

Design Considerations for a Message File Server

Abstract: In this paper we describe a message server facility for handling large organizational archives of messages in an office information system environment. Messages can be retrieved according to attribute values specified and to some pattern of words appearing in the text of the message. We discuss design decisions and performance considerations in this environment and we derive analytic formulas for the optimal choice of the parameters of the message file organization.

Queues in Which Customers Receive Simultaneous Service from a Random Number of Servers: A System Point Approach

Abstract: We examine a multi-server queueing system with Poisson arrivals in which customers require simultaneous service from a random number of servers. Servers assigned to the same customer begin and end service concurrently. Service times are, in general, assumed to be exponentially distributed. A system point approach is presented as a framework for obtaining the waiting time distribution for each customer type. Explicit solutions are derived for the two-server system.

The Berkeley Internet Name Domain Server

Abstract: The Berkeley Internet Name Domain (BIND) Server allows a standard way of naming the many types of objects and resources that exist in distributed UNIX environments, and provides operations for storing and retrieving information about these objects. BIND ServerS collectively manage a hierarchical name space that is partitioned into domains reflecting administrative entities. Many existing UNIX applications, particularly mail facilities, will benefit greatly from such a service.

The domain name system

Abstract: : The domain name system is a protocol and a set of servers which provides a uniform method for associating the names of resources (e.g., mailboxes, host names) to information about the resources (e.g. mail server addresses). The name database is distributed among multiple name servers scattered through one or more internets. The protocol provides tools for controlling both the distribution of the database and the responsibility for maintenance of the distributed pieces of the database. (Author)

Response Time Preservation: A general technique for developing approximate algorithms for queueing networks

Abstract: Response Time Preservation (RTP) is introduced as a general technique for developing approximate analysis procedures for queueing networks. The underlying idea is to replace a subsystem by an equivalent server whose response time in isolation equals that of the entire subsystem in isolation. The RTP based approximations, which belong to the class of decomposition approximations, can be viewed as a dual of the Norton's Theorem approach for solving queueing networks since it matches response times rather than throughputs. The generality of the RTP technique is illustrated by developing solution procedures for several important queueing systems which violate product form assumptions. Examples include FCFS servers with general service times, FCFS servers with different service times for multiple classes, priority scheduling, and distributed systems.

An M/G/c queue in which the number of servers required is random

Abstract: Many queueing situations such as computer, communications and emergency systems have the feature that customers may require service from several servers at the same time. They may thus be delayed until the required number of servers is available and servers may be idle when customers are waiting. We consider general server-completion-time distributions and derive approximation methods for the computation of the steady-state distribution of the number of customers in queue as well as the moments of the waiting-time distribution. Extensive computational results are reported.

Incorporating load dependent servers in approximate mean value analysis

Abstract: Queueing network performance modelling technology has made tremendous strides in recent years. Two of the most important developments in facilitating the modelling of large and complex systems are hierarchical modelling, in which a single load dependent server is used as a surrogate for a subsystem, and approximate mean value analysis, in which reliable approximate solutions of separable models are efficiently obtained. Unfortunately, there has been no successful marriage of these two developments; that is, existing algorithms for approximate mean value analysis do not accommodate load dependent servers reliably.This paper presents a successful technique for incorporating load dependent servers in approximate mean value analysis. We consider multiple class models in which the service rate of each load dependent server is a function of the queue length at that server. In other words, load dependent center k delivers “service units” at a total rate of f

A Queueing System with Auxiliary Servers

Abstract: We examine a queueing system with multiple primary servers and a fewer number of auxiliary servers. There are two classes of customers-those who require service from a primary server working alone and those who require service from a primary server who is assisted by an auxiliary server. Though the apparent Markovian state space is five-dimensional, we show that an aggregation results in an exact two-dimensional representation which is Markovian. Matrix geometric theory is used to obtain approximations for the mean delay and blocking probability of each customer type.

Technical Note—On Waiting Times for a Queue in Which Customers Require Simultaneous Service from a Random Number of Servers

Abstract: We consider a queueing system, first introduced by L. Green in 1980, in which customers from a Poisson arrival stream request simultaneous service from a random number of identical servers with exponential service times. Computational formulas for the second moment of time in queue are given, along with tables of these values for selected systems. Numerical results show that the coefficient of variation for time in queue is always greater than 1 and decreases with increasing congestion.

Linguistic Support of Receptionists for Shared Resources

Abstract: This paper addresses linguistic issues that arise in providing support for shared resources in large scale concurrent systems. Our work is based on the Actor Model of computation which unifies the lambda calculus, the sequential stored-program and the object-oriented models of computation. We show how receptionists can be used to regulate the use of shared resources by scheduling their access and providing protection against unauthorized or accidental access. A shared financial account is an example of the kind of resource that needs a receptionist. Issues involved in the implementation of scheduling policies for shared resources are also addressed. The modularity problems involved in implementing servers which multiplex the use of physical devices illustrate how delegation aids in the implementation of parallel problem solving systems for communities of actors.

Name Server Design for Distributed Systems.

Abstract: : Name servers are programs that provide a uniform method of naming resources and retrieving information about those resources in the context of a distributed system. This paper examines the main issues in name server design and considers some contemporary name servers in order to classify the common goals, methods, and tradeoffs. (Author)

Modeling file system organizations in a local area network environment

Abstract: This paper presents a modeling methodology for the design of file system organizations in a local area network environment. We first propose a characterization for the workload of typical business transaction systems, and then use this measurable characterization to derive input parameters for the file distribution graph models and the queuing network models that represent two completely different file system organizations, i.e., a file server-based file system and a distributed file system. Total system throughput and mean system response time are the indices used in comparing these two design approaches. An example of this comparison is given. It is shown that this methodology and these models are useful tools for the evaluation of certain design tradeoffs.

Performance of a DECnet based disk block server

Abstract: This report describes an experimental disk block server implemented for the RSX-11M Operating System using DECnet. The block server allows user programs on one system to access files on a disk physically located on a different system. The actual interface is at the level of physical blocks and IO transfers. Results of basic performance measurements are given, and explained in terms of major components. Performance predictions are made for servers of this type supporting more complex workloads.

Distributed match-making for processes in computer networks

Abstract: In the very large multiprocessor systems and, on a grander scale, computer networks now emerging, processes are not tied to fixed processors but run on processors taken from a pool of processors. Processors are released when a process dies, migrates or when the process crashes. In distributed operating systems using the service concept, processes can be clients asking for a service, servers giving a service or both. Establishing communication between a process asking for a service and a process giving that service, without centralized control in a distributed environment with mobile processes, constitutes the problem of distributed match-making. Logically, such a match-making phase precedes routing in store-and-forward computer networks of this type. Algorithms for distributed match-making are developed and their complexity is investigated in terms of message passes and in terms of storage needed. The theoretical limitations of distributed match-making are established, and the techniques are applied to several network topologies.

Modeling the Effect of Local Area Network Contention on the Performance of Host Computers

Abstract: Many computer systems have components with FCFS queueing and ser.vice lime distributions that may not be exponential. Exact models of such sys­ tems arc usually intractable. When tbe component is a local area network.gateway. the analysis is complicated by the possibility that a job's queueingdelay may depend nol only on the network'sservice pacameten. but also onthe load offered by all hosts attached 10 it. Modeling the delay at LAN gate­ways as self-contained units may be possible, but this does not teU us abouttheir impact on the performance of tbe system as a whole. A method ofpredicting this impact based on Marie's algorithm for the solution of elosedqueueing networks with Coxjan serve~ has been proposed. An open-queuemodel of a local area network is driven by the throughputs of eaeh of its gate­ways; the gateway at each host is treated as a Conan server whose queuelength will match that predicted by the open model. These Coxian servers arethcn used in Marie's algorithm to predict the pedormance of each bost inturn. The procedure is iterative; it is terminated when tbe convergence cri­teria for Maric'salgorithm are mel by the models of aU bosts_ The method. hasbcen applied to a modified version of the Berry-Chandymodel of a token ringwith noncxhaustive service. The results are close to those predicted by simula­lions of a ring with two hasls for various Iraffic loads.The choice of Marie's algorithm

AN MIGIc QUEUE IN WHICH THE NUMBER OF SERVERS REQUIRED IS

Abstract: Many queueing situations such as computer, communications and emergency systems have the feature that customers may require service from several servers at the same time. They may thus be delayed until the required number of servers is available and servers may be idle when customers are waiting. We consider general server-completion-time distributions and derive approximation methods for the computation of the steady-state distribution of the number of customers in queue as well as the moments of the waiting-time distribution. Extensive computational results are reported.

A pipeline and parallel architecture for supporting database management systems

Abstract: A pipeline and parallel architecture is described and analyzed for efficient support of database management systems. The main feature of the architecture is a RAM consisting of two bit-arrays, one word-array, and one memory bank. Through the use of the RAM, cross referencing involved in the join and project operations can be removed, without the need for sorting. Further performance improvements are achieved by using parallelism in tuple cancatenation and tuple comparison via a set of servers. The architecuture is quite suitable for a wide range of database applications, especially those requiring a high percentage of cross referencing. A hardware simulator was developed for determining the optimal number of servers for different applications.

The dependence of sojourn times on service times in tandem queues

Abstract: In this paper we study a series of servers with exponentially distributed service times. We find that the sojourn time of a customer at any server depends on the customer's past history only through the customer's interarrival time to that server. A method of calculating the conditional probabilities of sojourn times is developed.

The Design and Implementation of the Berkeley Internet Name Domain (BIND) Servers

Abstract: : Name servers are system processes that maintain databases of information about objects existing in computer networks, and answer user queries concerning them. Naming services are becoming increasingly important as computers are connected into networks and eventually into internetworks. This report describes our design and implementation of the DARPA Internet name servers in the Berkeley Unix environment. The naming service and the information it provides are both distributed and replicated user update queries, as well as retrieval queries, were implemented, and zones of network information are kept up to date by incremental zone refresh operations.

Some notes on the total service time in certain loss

Abstract: In this paper we discuss the total service rendered by all the servers in certain loss systems.

The electronic classroom: Workstations for teaching an extended abstract

Abstract: Continuing advances in hardware have made it possible to replace mainframe time-sharing systems (with their inherent performance limitations and poor user interfaces based on low-speed alphanumeric terminals) by powerful personal computers. When these personal computers have high-resolution bit-mapped graphics displays, fast processors, and virtual memory, and are linked in networks, they combine the best of dedicated computing (e.g., immediate response) with the best of time-sharing (e.g., resource-sharing of programs, data and peripherals). Personal computers so configured are typically called workstations. Such workstations have been introduced in the last few years primarily for professionals in productivity-sensitive areas like engineering design and office automation, but recent price/performance improvement has made it possible to consider them for use in education as well.Brown University has embarked on an ambitious program to populate the campus with high-performance graphics-based workstations for use in all disciplines. We project the installation of hundreds in the coming academic year and ramping up to several thousand in just a few years. By the end of the decade we expect to have up to ten thousand workstations, essentially one for every faculty, student, and administration/staff person who wants or needs one, as well as for affiliated personnel in hospitals, libraries and state government offices [Ship83]. These workstations are to be connected in local area networks (LANs) that in turn will be attached via gateways to a campus-wide backbone broadband (300MHz) network, BRUNET [Ship82], that has been fully operational since 1981. On one of its subchannels BRUNET carries packet-switched data between mainframes and their file and printer servers and LAN gateways.

Development of the NLTSS operating system

Abstract: The NLTSS implementation effort has been going on for some five years now. Many parts of the system have been running for several years, but are undergoing constant change. There are three areas to consider. First, the distributable network operating system and communications protocols require a careful study of requirements and goals. The development of the LINCS architecture in concert with the perceived structure of the NLTSS message system and servers was a complex social process in which we all had to learn the give and take of negotiation. Reaching consensus agreements was often a lengthy and uncomfortable process. Then choosing the algorithms for implementation and developing them in a network environment is complicated. Secondly, the implementation effort itself was complicated by the use of new programming techniques, e.g., use of the MODEL language, the design for co-servers. Determining efficient mechanisms for communication and input/output in a message passing network architecture takes time. Oftentimes, special case optimizations would be implemented too quickly. There was and still is a lot of try, learn, try again technique employed in the development of NLTSS and LINCS. In the beginning of the project all the various support tools were yet to be conceived andmore » implemented. These actually represent a major portion of the first couple of years' work.« less