作者:
Barbara, DanielGarcia-Molina, HectorPrinceton Univ
Dep of Electrical Engineering & Computer Science Princeton NJ USA Princeton Univ Dep of Electrical Engineering & Computer Science Princeton NJ USA
In a faulty distributed system, voting is commonly used to achieve mutual exclusion among groups of nodes. Each node is assigned a number of votes, and any group with a majority of votes can perform the critical opera...
详细信息
ISBN:
(纸本)0818605642
In a faulty distributed system, voting is commonly used to achieve mutual exclusion among groups of nodes. Each node is assigned a number of votes, and any group with a majority of votes can perform the critical operations. Vote assignments can have a significant impact on system reliability. To compare vote assignments, the authors define two deterministic metrics, node and edge, vulnerability. They present various properties of these metrics and discuss how they can be computed. For these metrics, they discuss the selection of the best assignment and propose heuristics to identify good candidate assignments. Finally, they summarize experimental results that indicate that the heuristics yield optimal or close to optimal assignments.
A technique is described for implementing k-resilient objects, i. e. , distributed objects that remain available, and whose operations are guaranteed to progress to completion, despite up to k site failures. The imple...
详细信息
ISBN:
(纸本)0818605642
A technique is described for implementing k-resilient objects, i. e. , distributed objects that remain available, and whose operations are guaranteed to progress to completion, despite up to k site failures. The implementation is derived from the object specification automatically, and does not require any information beyond what would be required for a nonresilient, nondistributed implementation. It is therefore unnecessary for an applications programmer to have knowledge of the complex protocols normally used to implement fault-tolerant objects. The technique is used in ISIS, a system being developed at Cornell to support resilient objects.
Several protocols that ensure that a database can be recovered to a consistent state after a transaction failure or system crash are compared. The study includes a collection of simple analytic models, based on Markov...
详细信息
ISBN:
(纸本)0818605642
Several protocols that ensure that a database can be recovered to a consistent state after a transaction failure or system crash are compared. The study includes a collection of simple analytic models, based on Markov processes, for these protocols and some surprising results on the relative performance of the protocols. The authors consider only two-stage transactions (all reads before writes) and ignore effects of serializing transactions. The most interesting performance result presented is that, for systems.obeying the assumptions of this paper, the 'pessimistic' policy of holding write locks to commit point is considerably less efficient than the 'optimistic' policy that allows reading of uncommitted data but risks cascading aborts. A multiversion policy was also studied and found always to be nearly as good as the optimistic policy and sometimes much better.
作者:
McKendry, Martin S.Georgia Inst of Technology
Sch of Information & Computer Science Atlanta GA USA Georgia Inst of Technology Sch of Information & Computer Science Atlanta GA USA
Examples to illustrate requirements for ordering mechanisms are introduced. A model of nested actions is then used as a basis for categorizing visibility requirements. These requirements go beyond those typical of dat...
详细信息
ISBN:
(纸本)0818605642
Examples to illustrate requirements for ordering mechanisms are introduced. A model of nested actions is then used as a basis for categorizing visibility requirements. These requirements go beyond those typical of databasesystems. because often the entities managed by operating systems.cannot be recovered if an action fails. Several simplifications that apply to many operating system problems are discussed. Algorithms for controlling ordering are then presented, with examples of their use. Several expediencies that result from ordering requirements are established. In many situations, recovery for nested actions can be implemented with a single backup copy of each item, a single synchronization variable can be used to control blocking, and generalized locking is not required. These savings appear to be fundamental to making the object-action approach viable for operating system construction.
作者:
Minoura, ToshimiOregon State Univ
Dep of Computer Science Corvallis OR USA Oregon State Univ Dep of Computer Science Corvallis OR USA
A typical database system maintains target data, which contain information useful for users, and access path data, which facilitate faster accesses to target data. Further, most large databasesystems.support concurre...
详细信息
ISBN:
(纸本)0818605642
A typical database system maintains target data, which contain information useful for users, and access path data, which facilitate faster accesses to target data. Further, most large databasesystems.support concurrent processing of multiple transactions. For a static database system model, where units of concurrency control are not dynamically created or deleted, various concurrency control methods are known. Also, many methods that allow concurrent accesses to indexing structures without invalidating their integrity are known. However, a straightforward integration of these two kinds of concurrency control methods fails because of the phantom problem. The author introduces group locks in order to solve this problem and discusses their implementation. It is shown that if the lowest-level access path data as well as the target data are two-phase locked by transactions, consistency of the logical data will be preserved.
For performance evaluation of distributedsystems.and concurrency control mechanisms, several concurrent tree algorithms with dynamic, distributed concurrency control have been proposed that combine selective locking ...
详细信息
ISBN:
(纸本)0818605642
For performance evaluation of distributedsystems.and concurrency control mechanisms, several concurrent tree algorithms with dynamic, distributed concurrency control have been proposed that combine selective locking policies and 'built-in' recovery mechanisms in place of more centralized concurrency control. Claims have been made that such algorithms allow a high degree of reliability and an improved degree of concurrency over traditional approaches. Because concurrency control mechanisms regulate the potential interference between concurrent processes, they have a dramatic effect on these aspects of system performance. An evaluation methodology that has been used to test these claims is described. The methodology combines evaluation of a prototype implementation of a concurrent system with simulation of the system on a more ambitious scale using a general-purpose multiprocessor simulation system. Results for the particular system on the degree of concurrency, degree of interference, and frequency of recovery generally supports the performance claims for the distributed concurrency-control mechanisms.
A technique is presented for estimating the performance of programs written for execution on fail-stop processors. It is based on modeling the program as a discrete-time Markov chain and then using z-transforms to der...
详细信息
ISBN:
(纸本)0818605642
A technique is presented for estimating the performance of programs written for execution on fail-stop processors. It is based on modeling the program as a discrete-time Markov chain and then using z-transforms to derive a probability distribution for time to completion. The use of z-transforms seems to have several potential advantages over the use of transition matrices, which is also discussed. Among these are the ease with which the transform associated with a program can be calculated, the simplicity of obtaining the mean and variance from the transform, and the promise of obtaining a closed form for the probability distribution. Nevertheless, the utility of the z-transform technique remains to be confirmed by its systematic application to realistic fault-tolerant programs.
作者:
Leitner, GeraldColumbia Univ
Dep of Computer Science New York NY USA Columbia Univ Dep of Computer Science New York NY USA
Many distributed applications are structured as a set of client processes generating requests for service, and a set of server processes servicing these requests. The common form of interprocess communication (IPC) be...
详细信息
ISBN:
(纸本)0818605642
Many distributed applications are structured as a set of client processes generating requests for service, and a set of server processes servicing these requests. The common form of interprocess communication (IPC) between two single processes does not provide adequate support for robustness and throughput optimization. The author presents a stylized form of IPC between a group of client processes and a group of server processes. The choice of a particular server process to service a request is made dynamically at run-time. This dynamic task distribution provides the kernel support needed to make individual failures of server processes transparent and to optimize system performance through load leveling. The semantics of stylized IPC and the high-level protocol to implement it are described. Particular emphasis is placed on minimizing the performance overhead incurred by this mechanism.
The symposiumproceedings contains 51 papers. Areas covered include: advanced system architectures;distributedsystems.methods and tools for software development;system applications in control and automation;conforman...
详细信息
ISBN:
(纸本)0444868461
The symposiumproceedings contains 51 papers. Areas covered include: advanced system architectures;distributedsystems.methods and tools for software development;system applications in control and automation;conformance testing of hardware and software;networking and data highways;system availability, reliability and maintainability. Twelve papers are in French, seven in German. Other papers are in English.
暂无评论