Engineering topics

Birman, Kenneth P.

Publications and source records attributed to Birman, Kenneth P..

At least 19 records

NASA NTRS1993

Process membership in asynchronous environments

The development of reliable distributed software is simplified by the ability to assume a fail-stop failure model. The emulation of such a model in an asynchronous distributed environment is discussed. The solution proposed, called Strong-GMP, can be supported through a highly efficient protocol, and was implemented as part of a distributed systems software project at Cornell University. The precise definition of the problem, the protocol, correctness proofs, and an analysis of costs are addressed.

Ricciardi, Aleta M.↗

NASA NTRS1993

The Isis project: Fault-tolerance in large distributed systems

This final status report covers activities of the Isis project during the first half of 1992. During the report period, the Isis effort has achieved a major milestone in its effort to redesign and reimplement the Isis system using Mach and Chorus as target operating system environments. In addition, we completed a number of publications that address issues raised in our prior work; some of these have recently appeared in print, while others are now being considered for publication in a variety of journals and conferences.

Birman, Kenneth P.↗

NASA NTRS1992

The process group approach to reliable distributed computing

The difficulty of developing reliable distribution software is an impediment to applying distributed computing technology in many settings. Experience with the ISIS system suggests that a structured approach based on virtually synchronous process groups yields systems that are substantially easier to develop, exploit sophisticated forms of cooperative computation, and achieve high reliability. Six years of research on ISIS, describing the model, its implementation challenges, and the types of applications to which ISIS has been applied are reviewed.

Birman, Kenneth P.↗

NASA NTRS1991

Design alternatives for process group membership and multicast

Process groups are a natural tool for distributed programming, and are increasingly important in distributed computing environments. However, there is little agreement on the most appropriate semantics for process group membership and group communication. These issues are of special importance in the Isis system, a toolkit for distributed programming. Isis supports several styles of process group, and a collection of group communication protocols spanning a range of atomicity and ordering properties. This flexibility makes Isis adaptable to a variety of applications, but is also a source of complexity that limits performance. This paper reports on a new architecture that arose from an effort to simplify Isis process group semantics. Our findings include a refined notion of how the clients of a group should be treated, what the properties of a multicast primitive should be when systems contain large numbers of overlapping groups, and a new construct called the casuality domain. As an illustration, we apply the architecture to the problem of converting processes into fault-tolerant process groups in a manner that is 'transparent' to other processes in the system.

Birman, Kenneth P.↗

NASA NTRS1991

Maintaining consistency in distributed systems

In systems designed as assemblies of independently developed components, concurrent access to data or data structures normally arises within individual programs, and is controlled using mutual exclusion constructs, such as semaphores and monitors. Where data is persistent and/or sets of operation are related to one another, transactions or linearizability may be more appropriate. Systems that incorporate cooperative styles of distributed execution often replicate or distribute data within groups of components. In these cases, group oriented consistency properties must be maintained, and tools based on the virtual synchrony execution model greatly simplify the task confronting an application developer. All three styles of distributed computing are likely to be seen in future systems - often, within the same application. This leads us to propose an integrated approach that permits applications that use virtual synchrony with concurrent objects that respect a linearizability constraint, and vice versa. Transactional subsystems are treated as a special case of linearizability.

Birman, Kenneth P.↗

NASA NTRS1991

The process group approach to reliable distributed computing

The difficulty of developing reliable distributed software is an impediment to applying distributed computing technology in many settings. Experience with the ISIS system suggests that a structured approach based on virtually synchronous process groups yields systems which are substantially easier to develop, fault-tolerance, and self-managing. Six years of research on ISIS are reviewed, describing the model, the types of applications to which ISIS was applied, and some of the reasoning that underlies a recent effort to redesign and reimplement ISIS as a much smaller, lightweight system.

Birman, Kenneth P.↗

NASA NTRS1991

Lightweight causal and atomic group multicast

The ISIS toolkit is a distributed programming environment based on support for virtually synchronous process groups and group communication. A suite of protocols is presented to support this model. The approach revolves around a multicast primitive, called CBCAST, which implements a fault-tolerant, causally ordered message delivery. This primitive can be used directly or extended into a totally ordered multicast primitive, called ABCAST. It normally delivers messages immediately upon reception, and imposes a space overhead proportional to the size of the groups to which the sender belongs, usually a small number. It is concluded that process groups and group communication can achieve performance and scaling comparable to that of a raw message transport layer. This finding contradicts the widespread concern that this style of distributed computing may be unacceptably costly.

Birman, Kenneth P.↗

NASA NTRS1991

Using process groups to implement failure detection in asynchronous environments

Agreement on the membership of a group of processes in a distributed system is a basic problem that arises in a wide range of applications. Such groups occur when a set of processes cooperate to perform some task, share memory, monitor one another, subdivide a computation, and so forth. The group membership problems is discussed as it relates to failure detection in asynchronous, distributed systems. A rigorous, formal specification for group membership is presented under this interpretation. A solution is then presented for this problem.

Ricciardi, Aleta M.↗

NASA NTRS1991

Programming with process groups: Group and multicast semantics

Process groups are a natural tool for distributed programming and are increasingly important in distributed computing environments. Discussed here is a new architecture that arose from an effort to simplify Isis process group semantics. The findings include a refined notion of how the clients of a group should be treated, what the properties of a multicast primitive should be when systems contain large numbers of overlapping groups, and a new construct called the causality domain. A system based on this architecture is now being implemented in collaboration with the Chorus and Mach projects.

Birman, Kenneth P.↗

NASA NTRS1990

The ISIS project: Fault-tolerance in large distributed systems

The semi-annual status report covers activities of the ISIS project during the second half of 1989. The project had several independent objectives: (1) At the level of the ISIS Toolkit, ISIS release V2.0 was completed, containing bypass communication protocols. Performance of the system is greatly enhanced by this change, but the initial software release is limited in some respects. (2) The Meta project focused on the definition of the Lomita programming language for specifying rules that monitor sensors for conditions of interest and triggering appropriate reactions. This design was completed, and implementation of Lomita is underway on the Meta 2.0 platform. (3) The Deceit file system effort completed a prototype. It is planned to make Deceit available for use in two hospital information systems. (4) A long-haul communication subsystem project was completed and can be used as part of ISIS. This effort resulted in tools for linking ISIS systems on different LANs together over long-haul communications lines. (5) Magic Lantern, a graphical tool for building application monitoring and control interfaces, is included as part of the general ISIS releases.

Birman, Kenneth P.↗

NASA NTRS1990

Tools for distributed application management

Distributed application management consists of monitoring and controlling an application as it executes in a distributed environment. It encompasses such activities as configuration, initialization, performance monitoring, resource scheduling, and failure response. The Meta system (a collection of tools for constructing distributed application management software) is described. Meta provides the mechanism, while the programmer specifies the policy for application management. The policy is manifested as a control program which is a soft real-time reactive program. The underlying application is instrumented with a variety of built-in and user-defined sensors and actuators. These define the interface between the control program and the application. The control program also has access to a database describing the structure of the application and the characteristics of its environment. Some of the more difficult problems for application management occur when preexisting, nondistributed programs are integrated into a distributed application for which they may not have been intended. Meta allows management functions to be retrofitted to such programs with a minimum of effort.

Marzullo, Keith↗

NASA NTRS1990

Fast causal multicast

A new protocol is presented that efficiently implements a reliable, causally ordered multicast primitive and is easily extended into a totally ordered one. Intended for use in the ISIS toolkit, it offers a way to bypass the most costly aspects of ISIS while benefiting from virtual synchrony. The facility scales with bounded overhead. Measured speedups of more than an order of magnitude were obtained when the protocol was implemented within ISIS. One conclusion is that systems such as ISIS can achieve performance competitive with the best existing multicast facilities - a finding contradicting the widespread concern that fault-tolerance may be unacceptably costly.

Birman, Kenneth P.↗

NASA NTRS1990

Tools for distributed application management

Distributed application management consists of monitoring and controlling an application as it executes in a distributed environment. It encompasses such activities as configuration, initialization, performance monitoring, resource scheduling, and failure response. The Meta system is described: a collection of tools for constructing distributed application management software. Meta provides the mechanism, while the programmer specifies the policy for application management. The policy is manifested as a control program which is a soft real time reactive program. The underlying application is instrumented with a variety of built-in and user defined sensors and actuators. These define the interface between the control program and the application. The control program also has access to a database describing the structure of the application and the characteristics of its environment. Some of the more difficult problems for application management occur when pre-existing, nondistributed programs are integrated into a distributed application for which they may not have been intended. Meta allows management functions to be retrofitted to such programs with a minimum of effort.

Marzullo, Keith↗

NASA NTRS1990

Low cost management of replicated data in fault-tolerant distributed systems

Many distributed systems replicate data for fault tolerance or availability. In such systems, a logical update on a data item results in a physical update on a number of copies. The synchronization and communication required to keep the copies of replicated data consistent introduce a delay when operations are performed. A technique is described that relaxes the usual degree of synchronization, permitting replicated data items to be updated concurrently with other operations, while at the same time ensuring that correctness is not violated. The additional concurrency thus obtained results in better response time when performing operations on replicated data. How this technique performs in conjunction with a roll-back and a roll-forward failure recovery mechanism is also discussed.

Joseph, Thomas A.↗

NASA NTRS1989

How robust are distributed systems

A distributed system is made up of large numbers of components operating asynchronously from one another and hence with imcomplete and inaccurate views of one another's state. Load fluctuations are common as new tasks arrive and active tasks terminate. Jointly, these aspects make it nearly impossible to arrive at detailed predictions for a system's behavior. It is important to the successful use of distributed systems in situations in which humans cannot provide the sorts of predictable realtime responsiveness of a computer, that the system be robust. The technology of today can too easily be affected by worn programs or by seemingly trivial mechanisms that, for example, can trigger stock market disasters. Inventors of a technology have an obligation to overcome flaws that can exact a human cost. A set of principles for guiding solutions to distributed computing problems is presented.

Birman, Kenneth P.↗

NASA NTRS1989

The role of order in distributed programs

The role of order in building distributed systems is discussed. It is the belief that a principle of event ordering underlies the wide range of operating systems mechanisms that were put forward for building robust distributed software. Stated concisely, this principle achieves correct distributed behavior by ordering classes of distributed events that conflict with one another. By focusing on order, simplified descriptions can be obtained and convincingly correct solutions to problems that might otherwise have looked extremely complex. Moreover, it is observed that there are a limited number of ways to obtain order, and that the choice made impacts greatly on performance.

Birman, Kenneth P.↗

NASA NTRS1989

Supporting large scale applications on networks of workstations

Distributed applications on networks of workstations are an increasingly common way to satisfy computing needs. However, existing mechanisms for distributed programming exhibit poor performance and reliability as application size increases. Extension of the ISIS distributed programming system to support large scale distributed applications by providing hierarchical process groups is discussed. Incorporation of hierarchy in the program structure and exploitation of this to limit the communication and storage required in any one component of the distributed system is examined.

Cooper, Robert↗

NASA NTRS1989

Exploiting replication in distributed systems

Techniques are examined for replicating data and execution in directly distributed systems: systems in which multiple processes interact directly with one another while continuously respecting constraints on their joint behavior. Directly distributed systems are often required to solve difficult problems, ranging from management of replicated data to dynamic reconfiguration in response to failures. It is shown that these problems reduce to more primitive, order-based consistency problems, which can be solved using primitives such as the reliable broadcast protocols. Moreover, given a system that implements reliable broadcast primitives, a flexible set of high-level tools can be provided for building a wide variety of directly distributed application programs.

Birman, Kenneth P.↗