Abstract: Research into the design of a “data manager” which is under the control of a network operating system is only beginning. The following critical aspect of this management is considered: file assignment within the network. Selected previous models and particularly their assumptions are given. An iterative numerical procedure is given which optimizes the file assignment in a computer network. It decomposes into a micro model for solving a queuing network problem and a macro model for solving an integer programming problem. The micro model determines the optimal branching probabilities to the hosts of the computer network from the current file assignment. It also introduces the concept of a real value for the degree of multiprogramming. The macro model reassigns the files to (possibly different) hosts such that the optimal branching probability constraints are satisfied. It also includes other constraints, such as the storage capacities of the hosts. The procedure is then repeated until convergence is achieved. Implications for dynamic file assignment are given, thus motivating a direction for future operating system design.

Abstract: A formal model for atomic commit protocols for a distributed database system is introduced. The model is used to prove existence results about resilient protocols for site failures that do not partition the network and then for partitioned networks. For site failures, a pessimistic recovery technique, called independent recovery, is introduced and the class of failures for which resilient protocols exist is identified. For partitioned networks, two cases are studied: the pessimistic case in which messages are lost, and the optimistic case in which no messages are lost. In all cases, fundamental limitations on the resiliency of protocols are derived.

Abstract: : This report documents the three year investigation by the National Bureau of Standards into the technical issues involved in providing a uniform user and program environment for (possibly concurrent) access to multiple heterogeneous remote database management systems. This report is an anthology of papers prepared by the investigators which identify and discuss the design, development and implementation of such a user interface, and the development of a demonstrable prototype. (Author)

Abstract: This report has been published in the Proceedings of the Fifth Berkeley Conference on Distributed Data Management and Computer Networks.

Abstract: This paper presents the architecture of a communications-oriented, real-time operating system named Squire. The Squire kernel provides memory management, preemptive multitasking, interprocess communication, and the ability to manage data outside the process address space, as well as services such as timers. User processes are protected from one another by means of restrictions on what objects they can access and on the type of access. Squire has been designed to provide efficient communication between cooperating processes, portability to new machine architectures, and support for multiple processor and distributed processor usage. Protection, reliability, and robustness have been major design goals. Squire supports a new kind of object called chunks, which exist outside the process address space, and can be used to store and manage data. Squire also supports a means for extending the kernel in a controlled manner; this mechanism is used both to implement such traditional functions as device drivers and to provide extended kernel services not present in the basic Squire kernel.

Abstract: A distributed system is modelled as a set of disjoint naming environments called contexts among which messages flow as the means of communication. A structured object is considered to be a list of the names of its components. Primitive objects contain no other objects as components. This paper presents an algorithm for copying objects across context boundaries. Each object resides within a context and has a locally unique name assigned to it. The components of a structured object may reside in other contexts. Three copy operations are proposed: (1) copy copies the complete structure without copying any component more than once; (2) copy-local copies all of the structure within the context of the object and directly accessible through the object without copying any component more than once, and sending to the receiver the globally unique names of those components not copied; (3) copy-top copies only the top level of the structure, sending to the receiver the globally unique names of all the components of the top level.