Walter H. Kohler

A Survey of Techniques for Synchronization and Recovery in Decentralized Computer Systems

Two related and fundamental problems in designing decentralized systems which support an object model of computation are introduced, and proposed solution techniques are surveyed The first problem is synchronizing access to shared objects while allowing a high degree of concurrency. The second is the recovery of objects in spite of user errors, application errors, or partial system failure. The synchronization problem is a general~.atmn of the concurrency control problem which arises in database and transaction-processing systems. Concurrency control methods which use locking, timestamps, mrculatmg permit, tickets, conflict analysis, and reservations are presented and compared. The proposed approach to solving the recovery problem is based on a software structuring abstraction called the atonuc action, a type of generalized transaction. Requirements and techniques for implementing atomic actions in a decentralized environment are discussed.

Characterization and Theoretical Comparison of Branch-and-Bound Algorithms for Permutation Problems

Branch-and-bound implicit enumeration algorithms for permutation problems (discrete optimization problems where the set of feasible solutions is the permutation group <italic>S<subscrpt>n</subscrpt></italic>) are characterized in terms of a sextuple (<italic>B<subscrpt>p</subscrpt> S,E,D,L,U</italic>), where (1) <italic>B<subscrpt>p</subscrpt></italic> is the branching rule for permutation problems, (2) <italic>S</italic> is the next node selection rule, (3) <italic>E</italic> is the set of node elimination rules, (4) <italic>D</italic> is the node dominance function, (5) <italic>L</italic> is the node lower-bound cost function, and (6) <italic>U</italic> is an upper-bound solution cost. A general algorithm based on this characterization is presented and the dependence of the computational requirements on the choice of algorithm parameters, <italic>S, E, D, L,</italic> and <italic>U</italic> is investigated theoretically. The results verify some intuitive notions but disprove others.

Debugging a Distributed Computing System

In this paper we discuss the issues involved in debugging a provide distributed computing system. We describe the major differences between debugging a distributed system and debugging a sequential program. We suggest a methodology for distributed debugging, and we propose various tools or aids.

A Preliminary Evaluation of the Critical Path Method for Scheduling Tasks on Multiprocessor Systems

The problem of scheduling tasks on a system of independent identical processors is discussed and the performance of a suboptimal method is evaluated. The computation is modeled by an acyclic directed graph G(T,<), where node set T represents the set of tasks to be completed and edge set < defines the precedence between tasks. The objective is to minimize the finishing time of the computation graph. Known theoretical results are reviewed and a general branch-and-bound algorithm for finding optimal solutions is presented. The schedules produced by a simple critical path priority method are shown to be near optimal for randomly generated computation graphs.

A modified priority based probe algorithm for distributed deadlock detection and resolution

A modified, priority-based probe algorithm for deadlock detection and resolution in distributed database system is presented. Various examples are used to show that the original priority-based algorithm, presented by M.K. Sinha and N. Natarajan (1985), either fails to detect deadlocks or reports deadlocks that do not exist in many situations. A modified algorithm that eliminates these problems is proposed. The algorithm has been tested through simulation and appears to be errorfree. The performance of the modified algorithm is briefly discussed. >

Concurrency features for the Trellis/Owl language

Trellis/Owl is an object oriented programming language being developed as part of the Trellis programming environment by the Object Based Systems group within Corporate Research and Architecture at Digital Equipment Corporations Hudson, MA facility. Trellis/Owl incorporates support for concurrency. Here we describe the high level language features designed, the rationale for that design, the implementation techniques used, the principles behind them, and experience with the resulting system to date. We show how the object oriented nature of Trellis/Owl influenced the design, and how it affected the implementation.

Exact, Approximate, and Guaranteed Accuracy Algorithms for the Flow-Shop Problem n/2/F / F

Improved exact and approximate algorithms for the n-job two-machine mean finishing time flow-shop problem, n/2JF/P, are presented While other researchers have used a variety of approximate methods to generate suboptimal solutions and branch-and-bound algorithms to gen- erate exact solutmns to sequencing problems, thin work demonstrates the computatmnal effectiveness of couphng the two methods to generate solutmns with a guaranteed accuracy. The computational reqmrements of exact, approximate, and guaranteed accuracy algorithms are compared expem- mentally on a set of test problems ranging in size from 10 to 50 jobs The approach is readily apphca- ble to other sequencing problems

Modeling the effects of data and resource contention on the performance of optimistic concurrency control protocols

The authors use a mean value model for data contention and a piecewise linear model for resource contention. To show the usefulness of this methodology, they compare three different optimistic concurrency control protocols for a centralized system. The authors derive a closed-form expression for the transaction throughput as a function of workload parameters and the resource-access-time parameters. The resource-access-time parameters can be derived using a simple analytical model. The closed-form expressions are very useful as a quick evaluation of different protocols and to gain insight about protocol performance over a wide range of model parameters. They also yield a simple asymptotic analysis of the optimistic concurrency control protocols. The authors apply the methodology to predict the performance of a testbed database system.<<ETX>>

Coupling small computers for performance enhancement

The advent of the microprocessor has opened up new avenues for the system designer to provide more powerful, more reliable and more user oriented computer systems to the user community for the same or lower costs. The problem confronting the designer is: How to achieve these goals? This paper describes one such method called Distributed Function Multiple Processor (DFMP). The system described uses several micro processors each with its own memory to form a cluster. These processors are differentiated by the functions they perform such as file managing, intelligent terminal, etc. and communicate via a Restricted Cross Bar Switch (RCBS). Further, several clusters or nodes can be linked to form a local network. Interprocessor and internode communications are controlled by a special processor called the Interprocessor Controller (PC) located in each node. The IPCs use an adaptive technique to determine traffic flows.

A queueing network model for a distributed database testbed system

A queueing network model to analyze the performance of a distributed database testbed system is developed and validated against empirical measurements. The model includes the effects of the concurrency control protocol (two-phase locking with distributed deadlock detection), the transaction recovery protocol (write-ahead-logging of before-images), and the commit protocol (centralized two-phase commit) used in the testbed system. The queueing model differs from previous analytical models in three major aspects. First, it is a model for a distributed transaction processing system. Second, it is more general and integrated than previous analytical models. Finally, it reflects a functioning distributed database testbed system and is validated against performance measurements.