No full-text available
To read the full-text of this research,
you can request a copy directly from the author.
Procédures de reprise dans les systèmes de gestion de base de données réparties
Abstract
In this paper, two recovery techniques used to restore the integrity of the distributed database management system operating in the presence of intermittent failures and with a checkpointing and roll-back recovery scheme are analysed and compared. A mathematical model of the distributed database management system is proposed to evaluate its disponibility at the updating and interrogating operations in the case of one, two, three and four distributed copies. Numerical illustrations are presented and it is shown that the recovery procedure based on the transfert of the valid copies across the network is better in some given conditions of processing of the distributed database management system.
An approximate model is presented for the mean response time in a distributed computer system in which components may fail. Each node in the system periodically performs a checkpoint, and also periodically tests the other nodes to determine whether they are failed or not. When a node fails, it distributes its workload to other nodes which appear to be operational, based on the results of its most recent test. An approximate response time model is developed, explicitly allowing for the delays caused by transactions being incorrectly transferred to failed nodes, because of out-of-date testing results. For the case when all nodes are identical, a closed form solution is derived for the optimal testing rate minimizing the average response time. Numerical results are presented illustrating the relationships among the problem parameters.
This paper presents a method for obtaining the optimum checkpoint interval of a transaction processing computer system subject to time dependent failures. The system uses checkpointing to create a valid system state, and roll-back in order to recover from failures. Maximizing system availability we derive the optimum checkpoint interval as a function of the load of the system and of the time dependent failure rate. The results are illustrated numerically for the Weibull failure rate.
A model for distributed systems with failing components is presented. Each node may fail and during its recovery the load is distributed to other nodes that are up. The model assumes periodic checkpointing for error recovery and testing of the status of other nodes for the distribution of load. We consider the availability of a node, which is the proportion of time a node is available for processing, as the performance measure. A methodology for optimizing the availability of a node with respect to the checkpointing and testing intervals is given. A decomposition approach that uses the steady-state flow balance condition to estimate the load at a node is proposed. Numerical examples are presented to demonstrate the usefulness of the technique. For the case in which all nodes are identical, closed form solutions are obtained.
Software systems in which many user's or programmers intervene may easily contain software items — such as viruses — which will endanger the integrity of the system. This paper proposes that in addition to the conventional recovery techniques, such as dumps and roll-back recovery, system availability be enhanced by the introduction of virus tests or other types of “failure tests”. We present a model to analyze the effect of the failure rate, the frequency of virus and failure testing, and the frequency of periodic dumps, on global system availability. We assume that the “failure” rate of the system increases as time elapses beyond any individual instant at which a virus test or failure test has been carried out. Thus, we are dealing with a system in which failures will be naturally time-dependent. We compute the optimum value of the interval between dumps, and also the best time interval between virus or failure tests for this system. In order to illustrate the methodology of this work, numerical examples are presented for various time-dependent failure statistics.
For a transaction oriented system, the authors propose that in
addition to the conventional recovery techniques, such as dumps and
roll-back recovery, system availability be enhanced by the introduction
of `failure tests'. They present a model to analyze the effect of the
failure rate, the failure tests, and the periodic dumps, on global
system availability. They then compute the optimum value of the interval
between dumps, and also the best time interval between failure tests for
this system. Numerical examples are presented for various failure models
One of the basic problems related to the efficient and secure operation of a transaction oriented file or database system is the choice of the checkpoint interval. It is shown that the optimum checkpoint interval (i. e. the time interval between successive checkpoints which maximizes system availability) is a function of the load of the system. It is proved that the total operating time of the system (and not the total real time) between successive checkpoints should be a deterministic quantity in order to maximize the availability. An explicit expression for this time interval is obtained. The results are a significant departure from previous work where load independent results have been obtained. A rigorous analysis of the queueing process related to the requests for transaction processing arriving at the system is also presented, and the ergodicity conditions for the system are proved.
A model is developed for allocating information files required in common by several computers. The model considers storage cost, transmission cost, file lengths, and request rates, as well as updating rates of files, the maximum allowable expected access times to files at each computer, and the storage capacity of each computer. The criterion of optimality is minimal overall operating costs (storage and transmission). The model is formulated into a nonlinear integer zero-one programming problem, which may be reduced to a linear zero-one programming problem. A simple example is given to illustrate the model.
The goal of this research is the solution of several problems involved in the optimal design of computerized message processing and communication systems. The class of systems studied includes airline and hotel reservation systems, time-shared computer systems, and corporate message communication systems. A precise mathematical model of the systems being studied is formulated to draw attention to the fundamental subsystems and to clarify the basic independent variables of the design processes. Systems considered must have a large data base organized into record files, widely distributed users, an on-line communication network joining users to data base sites a quantitative performance measure and specified performance constraints. The most general system model studied uses a queueing model for individual communication channels, with the channel interconnections specified by a weighted linear graph. One major problem area studied is the determination of the optimal number and locations of sites for the system files. A second major problem considered is the optimal design of communication channels and networks of these channels. The third major problem area studied is the optimal design of communication network topologies. (Author)
A mathematical model of a transaction-oriented system under intermittent failures is proposed. The system is assumed to operate with a checkpointing and rollback/recovery method to ensure reliable information processing. The model is used to derive the principal performance measures, including availability, response time, and the system saturation point.
The advent of computerized information networks connotes a change in the normal data processing concept of information as a localized object, a disk file tied to computer X, for instance. Having available a facility for communication between computers it is more natural to think of related data files at different sites as elements of a single unified data collection. As an example, crime reports gathered in various cities across the nation inherently constitute an integrated data base of more than local interest. In banking, manufacturing, and other applications the same is often found to be true; information gathered and maintained in one geographical area is closely related to data collected elsewhere. Computer networks offer not only a replacement for unwieldy document-oriented methods of communicating between distributed data files, but also promise the flexibility needed to develop new ways of applying these data collections.
We consider a mathematical model of an information network of n nodes, some of which contain copies of a given data file. Within this network, every node is able to communicate with every other node over communication links (a process which may entail routing through intermediate nodes). In particular, we are concerned with transactions with the multiply-located file. Such transactions fall into one of two classes: (1) query traffic between a node and the file, and (2) update traffic. An update message is assumed to be transmitted to every copy of the file, whereas a query is communicated only to a single copy.
A stochastic model of a transaction oriented computer system in the presence of intermittent failures, operating with a checkpoint and roll-back recovery scheme, is proposed in the case of a hierarchy of checkpoints and failures. An analysis provides the stationary probability distribution for the model and the optimum checkpoint intervals for two cases of interest: when a fixed checkpoint is used for each failure and when the allowable checkpoint closest in time is used.
The interaction between computers and communication technology has steadily developed in recent years bringing with it the creation of large computer communication complexes. Earlier computer communication systems were generally focused around a single large computer installation. Although a strong argument can still be made for serving distributed users with a centralized system, we now witness an increasing number of systems in which information processing and storage functions are distributed among several computers. The idea is to distribute the resources (programs, data, computing hardware, etc.) of each computer site to a widely spread community of users. Various factors may favor a shift toward computer networks, especially with similarly structured organizations which are independent but can be motivated to share. The most well-known example is the network developed under the sponsorship of the Advanced Research Project Agency (ARPA), that links independent research organizations, all involved in solving closely related problems. The libraries of specialized software and data at each research center, make it attractive to link the computers of these organizations. Clearly, access to specialized data bases is a major feature of the networks and the cost incurred by such an access is an important consideration in determining the economic viability of this kind of resource sharing.
Three classes of file directories for distributed data bases are studied: The centralized directory system, the local directory system, and the distributed directory system. The parameters considered are communication cost, storage cost, code translation cost, query rate, update rate, directory size, and directory response time. This study reports the cost performance tradeoffs of these three classes of directory systems in star and distributed networks and provides a guide in the design of file directory systems when operating in these network environments.
The problems of file allocation and capacity assignment in a fixed topology distributed computer network are examined. These two aspects of the design are tightly coupled through an average message delay constraint. The objective is to allocate copies of information files to network nodes and capacities to network links so that a minimum cost is achieved subject to network delay and file availability constraints. A model for solving the problem is formulated and the resulting optimization problem is shown to fall into a class of non-linear integer programming problems. Deterministic techniques for solving this class of problems are computationally cumbersome even for small sized problems. A new heuristic algorithm is developed, based on a decomposition technique which greatly reduces the computational complexity of the problem. Numerical results for a variety of network configurations indicate that the heuristic algorithm, while not theoretically convergent, yields practicable low cost solutions with substantial savings in computer processing time and storage requirements. Moreover, it is shown that this algorithm is capable of solving realistic network problems whose solution using deterministic techniques is computationally intractable.
The problems of file allocation and capacity assignment in a fixed topology distributed computer network are examined. These two aspects of the design are tightly coupled by means of an average message delay constraint. The objective is to allocate copies of information files to network nodes and capacities to network links so that a minimum cost is achieved subject to network delay and file availability constraints. A model for solving the problem is formulated and the resulting optimization problem is shown to fall into a class of nonlinear integer programming problems. Deterministic techniques for solving this class of problems are computationally cumbersome, even for small size problems. A new heuristic algorithm is developed, which is based on a decomposition technique that greatly reduces the computational complexity of the problem. Numerical results for a variety of network configurations indicate that the heuristic algorithm, while not theoretically convergent, yields practicable low cost solutions with substantial savings in computer processing time and storage requirements. Moreover, it is shown that this algorithm is capable of solving realistic network problems whose solutions using deterministic techniques are computationally intractable.
A study of Optimal File Assignment and Communication Network Con-figuration
- V K M Whitney
Whitney, V.K.M.: A study of Optimal File Assignment and Communication Network Con-figuration. Ph. D. Dissertation, University of Michigan, 1970 Re~u le 11 mai/12 d6cembre 1978
The Distributed Control of Multiple Copies of Data
- A P Mullery
Cohérence et gestion d'objets dupliqués dans les systèmes distribués
- J Seguin
- G Sergeant
- P Wilms