J. Michael Bennett

Characteristics of files in NFS environments

As local area networks of workstations and file servers grow, the management of user and system files becomes increasingly more challenging for system administrators. In order to develop effective management tools for these administrators, the nature of files in workstation environments must be understood. This paper presents a static analysis of file usage patterns of a typical workstation environment using a standard client-server paradigm (NFS). The data reveals some differences in characteristics for files in networked environments compared with files in timesharing systems. The study also suggests what data and techniques might be appropriate as a basis for a management tool for network administrators. The paper concludes with a recommendation for the construction of a real-time management and data gathering tool that may give early-warnings of server overload or may be useful in exploring trade-offs among file allocation strategies.

A resource management system based on the ODP trader concepts and X.500

Distributed computing systems are composed of various types of hardware and software resources. Providing a reliable and efficient distributed computing environment largely depends on the effective management of these resources and the services that they provide. ISO has begun work on a proposed standard for Open Distributed Processing (ODP). The ODP framework includes a mechanism called the Trader which provides a framework for exchanging services in an open distributed computing environment. This paper presents a design of a resource information management system which employs and extends the ODP Trader concepts to facilitate the management and use of resources, information about resources and the services provided by the resources. We describe the architecture, information model, and user interface of the resource management system. We also describe a prototype implementation which uses the X.500 Directory Service as its repository for resource information and report on our experience with it to date.

Distributed search for cooperative applications

Applications at individual sites throughout a network can cooperate to find information re uested by a user. % These apphcations cooperate by exc angmg mqsaa es ? and are autonomous except that all have been Imp emented to respond to a common protocol. Every ap li‘F cation 1s responsible for mamtamlng Its own local m ormation. Further, each application haa limited knowledge of the other sites in the network, namely, those siteq i~med!ately connected to lt which run a, cooperating application. Thw collection of apphcatlons 1s referred to as a (loosely) distributed application. The emerging standard on distributed directories (X.500) is an example of such a distributed application. A distributed search algorithm is presented which is used by each ap Iication to initiate a se,arch or to ref spend to a searc request from another site. The cooperating applications are guaranteed to find and ~eturn an appropriate ~eply. to the ori inating site. It 1s also # shown that no site w1ll retain m o~matlon about a message for an indefinite amount of time. “Support for this research was provided in part by the National Science and Research Council of Canada. ACM-SAC ‘931WMIUUSA O 1993ACMCM9791—%&Z@~l..M~ Introduction The Iar e scale interconnection of many diffe~ent hosts J has m e applications, such as electromc mad and file transfer important tools m many different organizations. ~his large scale connectivity and diversity of hosts has also made it difficult for a user. to identify individuals or hosts of Interest without prior knowledge. This problem has motivated research mto the use of online sources of knowledge about users, hosts, and other objects within electronic mail systems[4, 6], work on attribute-baaed naming[l] and work on cooperating information sources[5, 7]. Such problems have motivated the work on X.500 [2] to support telecommunication applications such as electronic mail (X .400[3, 8]) and other telematic services (e.g., FAX, Telex). The emerging X.500 standard introduces the notion of a dtstm”buted directory ~ a single, global application distributed across many different, autonomous .sItes.. The standard specifies {rotoco’s ‘0 ~y directing queries to local directory agenLs. These ermlt other apphcatlons to make use of t e directory agents cooperate ; F “ rovide the designated directory serwces t , the% a p Icatlons. The dlrector~ information itself IS dlstrl uted throughout these sttes and is managed by the directory agents. On oing work on the f X.500 standard IS focu=d on protoco s and procedures for dwtrlbutmg and replicating directory information. A similar scenario has been reposed for networks of library systems. A user woul~ from his or her own home library, make a request for a particular book.. The distributed mfo~matlon agents would cooperate m Iocatin a hbrary which had a copy of the desired text. Suc 1! an information network could even tap into publishers’ databases (with suitable protection) and inform the user of an ordering source or publication date. Although these applications may differ in many ways, there 1s a common problem at the core of then behawour the retrieval of information from one or more sources dwtrlbuted throughout a distributed system. This problem has the following characteristics: . The applications are essentially autonomous except

On the use of distributed joins for processing interlibrary loans

This paper presents some experimental results in applying standard distributed jcin algorithms to a common practical problem in the library community; the efficient processing of interlibrary loans. We report on the performance (through simulation) of a number of different join algorithms and show that local join processing methods such as indezed nested-loop and hash-join, with the addition of specific attribute indices, are more cost-effective in minimizing communication and local processing costs than standard methods such as sort-merge. These results axe applicable to a wide range of distributed database applications in addition to library databases.

Experiences in the establishment of a microcomputer support laboratory

The explosive growth and general availability of microcomputers is creating a desperate need for the training of computer programmers in how to approach, assess and, if applicable, use microcomputers as an addition to their traditional hardware/software tool-kit in solving applications problems. Until recently, our students (and, we suspect, CS majors and software engineers in general) had little or no formal training or experience in this area. One cannot appreciate the full extent of the applicability and limitations of micros without reasonable hands-on experience. Yet the obvious approach - providing a number of stand-alone micros as the sole teaching vehicle - of itself creates a misleading context: restrictions of every kind force the student to spend most of his time circumventing limited resources and far too little time in actually learning about the useful aspects of the machines per se and as components of a more complex whole. In this paper, we define the educational goals that we are trying to achieve. Then we identify five possible solutions to the problem. We then analyze each and present reasons why we adopted the approach we did. Our Microcomputer Support Laboratory [MSL], consists of a PDP-11/34, under UNIX@@@@, connected to 8 Motorola 6809s and 8 MC68000s. We address also the reasons which lead to this choice, the problems in building the MSL, practical experiences derived from using it in an educational situation, and conclude by discussing the feasibility of providing this type of environment as a university-wide resource. We also indicate possible improvements and future directions in which we hope to grow.

Integrating the personal computer into the university computing center environment: “A marriage of necessity!”

In this paper an approach to the integration of personal computers with a university computing center is outlined. The approach involves the development of application-oriented software local to the personal computer effectively transforming it into a software workstation. This permits users to take advantage of the features and resources of both the personal computer and the main-frame. To facilitate the development of such software the paper outlines an approach to the design of workstations. An experiment involving the development of a simple editing workstation based upon this method is described and discussed.