Abstract: Citation for published version (APA): Hooman, J. J. M., & Roever, de, W. P. (1990). Design and verification in real-time distributed computing : an introduction to compositional methods. In E. Brinksma, G. Scollo, & C. A. Vissers (Eds.), Protocol Specification, Testing and Verification IX (Proceedings of the 9th IFIP WG 6.1 International Symposium, Enschede, The Netherlands, June 6-9, 1989) (pp. 37-56). North-Holland Publishing Company.

Abstract: The Cambridge Distributed Computing System (CDCS) was designed some ten years ago and was in everyday use at the Computer Laboratory until December 1988. An overview of the basic design of CDCS is given, an outline of its evolution and a description of the distributed systems research projects that were based on it. Experience has shown that a design based on a processor bank leads to a flexible and extensible distributed system.

Abstract: The distributed global coin flipping problem is to combine the private random source s of individual processors to obtain a single source of randomness visible to all processors i n the system. Attention was focussed on global coin flipping in 1983, when Michael Rabin showed that, given a method of flipping a global coin in constant time, Byzantine agreemen t can be achieved with n processors, of which t < n/4 may be faulty, in an expected 0(1 ) rounds of communication. This was a substantial improvement over the worst case runnin g time for deterministic algorithms of t + 1 rounds . The first proposed algorithm for globa l coin flipping, suggested by Andrei Broder and Danny Dolev in 1984, required an encryption scheme with several conditions . Johan Hastad showed these conditions were not satisfied b y RSA with low exponent . Hastad's work marked a turning point, after which research on th e applications of cryptography to distributed computing increased sharply . Consider the following naive (and incorrect) solution to the global coin flipping problem . For simplicity, assume the existence of a broadcast channel . Each participant p i chooses a random bit bi and a random encryption/decryption pair of functions (Ei , D i ), and broadcast s Ei and Ei (bi ) on the broadcast channel . This commits each p i to its chosen bit b i . After al l participants have committed, each p i reveals b i by publishing the decryption key D i . The b i 's are then summed modulo 2 and the resulting bit is the global coin . Let p c be correct and P f faulty. While P f cannot guess b, from Ee (b c), by blindly "replaying " pc 's messages p f can publish an encryption of b, : p f simply chooses Ef = E~ and sets E f (bf) = E,(b,) . (During decryption p f could again "replay" p c 's message, setting D f = D c .) More generally, for an arbi trary function F it may be possible, given Ei„ EX27 . . . , Ei . and E i, (b i , ), Ei2 (b it ), . . . , Ei ,. (b,,, ) , to efficiently produce an encryption key E and E(F(b i,,bi27 . . . , bik )) (and compute the de cryption function for E from D i , , D2 2 , . . . , D ik ) . Even if the individual bits b i are created and encrypted completely according to protocol , there is no way to force a faulty pf to publish its decryption key Df . Thus, if p f doesn' t like the way the coin flip is turning out, it could refuse to decrypt its bit, thereby biasin g the outcome. To prevent this a method is needed for verifiable secret sharing (VSS) . In this problem, first posed and solved by Chor, Goldwasser, Micali, and Awerbuch, a dealer p distributes a secret m among the participants in such a way that (1) if p is correct during distribution and not too many participants fail, then m can be recovered, even if p itself fails during reconstruction and (2) if p is not correct during distribution then either al l correct participants will discover its faultiness during distribution or p will be committed t o a message m and during reconstruction m will be the value obtained for p, despite the actions of faulty participants . Moreover, (3) if p is correct during distribution and the number of faults does not exceed a certain threshold t, then no t faulty players learn any information about the secret m until reconstruction .

Abstract: The authors examine the use of grids that are conformal with the geometry of the material boundaries and, consequently, reduce the discretization errors introduced by the stair-stepped approximation. However, it is found that the introduction of the nonuniformity in the grid gives rise to at least two difficulties in the implementation of the finite-difference time-domain algorithm. The first is the loss of accuracy in the computation of finite-difference derivatives when the field points are distributed nonuniformly. The second is implementation of the appropriate boundary conditions at the material interfaces. The authors address these problems and suggest some means for eradicating them. A scheme is presented that has the attractive feature that it reduces to the conventional finite-difference time-domain method if a uniform grid is used. To illustrate the proposed algorithm the authors consider a parallel plate waveguide which supports a single propagating mode. >

Abstract: An unstructured grid flow solver was implemented on a massively parallel computer, and benchmark computations were performed. The solver was a two-dimensional computational fluid dynamics (CFD) code that performs first-order, steady-state solutions of the Euler equations. The parallel computer employed was the Connection Machine made by Thinking Machines, Corp. The CFD code was programmed in *Lisp, the accuracy of the code was verified, and numerous optimizations were implemented. Several benchmark runs were then made to assess and understand the impact of the code modifications and to obtain meaningful performance comparisons with other advanced computers.

Abstract: A description of an automatic two-dimensional finite-element grid generator which greatly assists in improving preprocessing operations and geometry handling for electromagnetic field computations is presented. The algorithm developed iteratively triangulates a set of points, defining a field region, into optimal finite-element meshes. The proposed mesh generator uses an approach proven to be user-friendly and efficient, especially in analysis of medical devices. This is because instead of using node-insertion techniques, as employed in most grid generators, the user gives dimensions and characteristics of each subregion in the large region to be analyzed. Furthermore, it is possible to add, delete, or modify a region configuration and grid without gross efforts. The triangulation and mesh refinements are performed automatically by the software without user intervention. The mathematical procedure used in developing the algorithm is based on the concept of Delauney triangulation. Variational principles are used to solve the electromagnetic field problem. Several examples are presented to show the effectiveness of this grid generator in handling a variety of problems and geometrical configurations. >

Abstract: Networking technology is in an era of accelerated growth. The availability of international and de facto standards is leading to unprecedented inter-connectivity among systems from multiple vendors. Innovative silicon technology has dramatically reduced the cost of physical connections to high-speed local area networks. Personal processing power is benefiting from extra CPU resources that are routinely being added, improving application and network usage. Special-purpose mini-supercomputers can handle compute- and I/O-intensive tasks at a fraction of the cost of yesterday’s supercomputers.

Abstract: The grid file can be used for the implementation of a database to handle range queries and partially specified queries. Although an upper bound of two disc accesses for a single record retrieval is guaranteed, the number of directory entries is an exponential function of the number of data buckets if the data distribution is heavily non uniform. A new method is presented for partitioning the search space for constructing a grid directory. It improves the ratio of the number of directory entries to the number of data buckets. The directory expansion appears to have a quadratic or cubic rate when the data distribution is non uniform. An implementation of the method is presented. Simulation results are given for both the traditional grid file partitioning method and the new method. >

Abstract: The DAPHNE system is a collection of tools and run-time support creating a network operating system interface for distributed programming using Modula-2. DAPHNE accomodates hardware and operating system heterogeneity without necessarily unifying heterogeneous system features. In describing the essentials of the DAPHNE approach we concentrate on the semantics of heterogeneous distributed programs and on the difficulties encountered in implementing DAPHNE on top of standard operating systems. Several requirements for the functionality of system interfaces are derived.

Abstract: The author describes the application of computer-generated images (CGI) in training and terrain analysis devices. Many of these images are generated by converting a digital grid map (x, y, z) grid database to a perspective view which is then displayed on a TV monitor. The author presents a unique mathematical approach for this conversion and a basic functional design for real-time (TV-rate) display of these images. A seascape image generated is also described where the database is a mathematical model. The proposed techniques have been applied to generate very realistic terrain and seascape images. Pseudonoise patterns have been used to enhance realism. It is concluded that this has been very effective in the foreground where one x/y grid element covers many pixels. The approach is limited to about a 20 degrees pitch down of the boresight. The computations required are primarily a series of accumulations. >

Abstract: The authors propose models for defining the availability metric of a distributed computing environment. The metric is a meaningful indicator of a distributed computing environment's performance from the reliability point of view. The proposed models can also be used to efficiently design, engineer, and monitor a distributed computing environment's activities. The models are general and can be applied to any distributed computing environment for availability evaluation. >

Abstract: This short article presents a part of our research results in object-based distributed computing: simulation results of optimistic and pessimistic synchronization mechanisms in distributed computing environments, where communication latency is not small. We show, for some difficult problems in distributed computing, e.g., distributed simulation of systems with tight feed back loops, any approach based on either sole optimistic synchronization or sole pessimistic synchronization does not work efficiently. For the purposes of highly efficient execution, we have to predict the degree of communication latency and the average behavior of the system and combine both synchronization mechanisms in an proper manner based on the prediction.

Abstract: End users can be classified into six distinct types. Each of them needs differentiated education, support, and control from the Information Systems function. To support a large number of their applications a new computing environment, “the third environment” must be developed by Information Systems (I/S) management. Close attention must also be paid by I/S management to the need to involve “functional support personnel” (end users in each functional area who spend most of their time programming and aiding other end users) in the I/S end user management process.