Shojiro Muro

Evaluation of the File Redundancy in Distributed Database Systems

This paper treats the file redundancy issue in distributed database systems, asking what is the optimal number of file copies, given the ratio r of the frequency of update requests to the frequency of all file access requests (i.e., queries and updates). Formulations of this type of problem, including optimal file allocation, have been attempted by a number of authors, and some algorithms have been proposed. Although such algorithms can be used to solve particular problems, it seems difficult to draw general conclusions applicable to a wide variety of practical distributed database systems. To probe into this hard to formulate but interesting problem, our paper constructs simplified network models of distributed database systems, and computes the optimal number of file copies, as well as their locations, to minimize the communication cost. For several network types, we plot the optimal number of file copies as a function of the ratio r.

Multi-version concurrency control scheme for a database system

Abstract A concurrency control scheme using multiple versions of data objects is presented which allows increased concurrency. The scheme grants an appropriate version to each read request. Transactions issuing write requests which might destroy database integrity are aborted. It is precisely stated when old versions can be discarded and how to eliminate the effects of aborted transactions is described in detail. The scheduler outputs only (multi-version) ww-serializable histories which preserve database consistency. It is shown that any “D-serializable” history of Papadimitriou (J. Assoc. Comput. Mach. 26 (4) (1979), 631–653) (or “conflict-preserving serializable log” of Bernstein et al., IEEE Trans. Software Engrg. SE-5 (3) (1979), 203–216) is ww-serializable.

Using branch-and-bound algorithms to obtain suboptimal solutions

AbstractThe branch-and-bound principle is successful in solving various combinatorial optimization problems. In general, however, the computation time becomes excessive as the sizes of problems grow. To overcome this difficulty the following three suboptimal methods are often employed in practice.(1)Letz be the value of currently known best solution of a given minimization problem, and letg (Pi) be a lower bound on the objective value of partial problemPi. TerminatePi ifg (Pi) ⩾z−∃ (z), where∃ (z) is an allowance specified in advance.(2)Cut off the computation as soon as T0 partial problems are decomposed, where T0 is a prespecified positive integer.(3)Always take into account at mostM0 (a given positive integer) number of active partial problems. The overflown partial problems are simply ignored. The effects of these methods on the computation time and the quality of obtained suboptimal solutions are investigated from both theoretical and simulation points of view.ZusammenfassungDas Branch-and-Bound-Prinzip kann erfolgreich zur Lösung verschiedener kombinatorischer Optimierungsprobleme eingesetzt werden. In der Regel nimmt der Rechenaufwand jedoch sehr stark mit wachsender Problemgröße zu. Um dieser Schwierigkeit zu begegnen, werden folgende drei suboptimale Methoden oft in der Praxis angewendet:(1)Seien z der Zielfunktionswert der besten bekannten Lösung eines gegebenen Minimierungsproblems undg (Pi) eine untere Schranke für die Zielfunktion des Teilproblems Pi. Dann terminierePi, fallsg (Pi) ⩾z−∃ (z) ist, wobei∃ (z) eine vorgegebene Fehlerschranke ist.(2)Beende die Rechnung, sobald T0 Teilprobleme bearbeitet worden sind, wobei T0 vorgegeben ist (Beschränkung der Rechenzeit).(3)Berücksichtige stets höchstensM0 (eine vorgegebene Zahl) aktive Teilprobleme. Die übrigen Teilprobleme weiden ignoriert (Beschränkung des Speicherplatzes). Die Auswirkungen dieser drei Methoden auf die benötigte Rechenzeit und die erhaltenen suboptimalen Lösungen werden theoretisch und an Hand von Simulationen untersucht.

Performance Analysis Of A CSMA/CD With Prioritized Access Waiting Time

We propose a new CSMA/CD protocol with priority called PW-CSMA/CD. In the PW-CSMA/CD, the prioritized scheme is realized by introducing the basic access waiting time such that the higher the priority level of message packets, the shorter the access waiting time. Characteristics of the PWCSMA/CD such as the average throughput and the number of times of retransmission, are theoretically analyzed. Compared to the original CSMA/CD, the obtained results for the PW-CSMA/CD show that the number of times of retransmission of highly prioritized packets is remarkably reduced for every t hroughput value and is also reduced to some extent for the case of packets with low priority. The mechanism and performance of the PW-CSMA/CD are further compared with those of the Priority Ethernet which is another CSMA/CD protocol with priority.

Throughput analysis of reservation protocols with tree-type reservation channel

Numerous multiple access protocols have been proposed aiming at an efficient utilization of transmission channels. Among those, the tree-type protocol is one of the multiple access protocols which provide a stable throughput. Various studies have been made on the tree-type communication protocols. This paper is concerned especially with the analysis of the throughput of the reservation protocols with tree-type reservation channel. In the proposed system, the tree-type protocol is employed to reserve the data slot, providing stable and high throughput. There are two types of tree-type protocols: the blocked-access tree protocol which inhibits the transmission of a new packet in the collision resolution interval; and the free-access tree protocol which does not inhibit the transmission. The reservation protocols employing these two protocols are discussed. Through the analysis, the optimum frame structure for each type is obtained. The maximum throughput of CCA proposed by Lee and Mark, as well as the optimum number of the small reservation slots, are obtained analytically.

Throughput analysis of Tree-Type protocols

Tree-type protocols achieve stable throughput even under heavy channel traffic compared with random access protocols such as the ALOHA types which show bistable behavior. Various binary tree-type protocols have been studied extensively for channel stability and the improvement of throughput performance. In this paper, closed-form solutions of the collision resolution time of several d-ary tree type protocols are obtained, and the effect of the degree value of their protocols on the throughput performance is discussed. As d-ary tree-type protocols, we consider the tree protocol, improved tree protocol, adaptive tree protocol and tree protocol with control mini-slot, and compare their performance. Further, by comparing the stability of the binary tree protocol with that of three ALOHA schemes with different backoff algorithms, we show that the binary tree protocol exhibits very stable behavior even for a heavy traffic load.

Performance analysis of the acknowledging ethernet

A modified Ethernet designated “Acknowledging Ethernet,” which essentially provides higher reliability, was proposed by Tokoro et al. They evaluated its performance by means of simulation. In this paper, the performance of the Acknowledging Ethernet is analyzed theoretically. The average throughput, the average number of times of retransmission, and the average data packet transmission delay are approximately obtained. The accuracy of the approximation analysis is examined by comparison with the simulation results. It was found also that the performance of the Ethernet is only slightly degraded with the introduction of the acknowledging mechanism in the Acknowledging Ethernet.

Multiversion Concurrency Control Scheme for a Distributed Database System - A Trial to Break Concurrent Update of Redundant Copies

A new multiversion concurrency control scheme for a distributed database system is proposed in this paper. Each data object in our database model has two copies allocated in different sites in the system. Unlike the usual distributed database systems with redundant copies, these two copies are not concurrently updated, but only one is updated for a write request. For each data object, the copy with a newly updated value is called the new version and the other is called the old version. Since two versions are accessible for each read request to an object and concurrent update of two copies of each data object is not necessary, our scheme allows increased concurrency. Our concurrency control scheme employs both timestamp mechanism and locking mechanism with two different modes which is based on the (r,a,c)-locking protocol proposed by Bayer et al., and it grants a version for every read request without causing inconsistency. Transactions with write requests which would cause inconsistency are aborted. It has been proven that our concurrency control scheme works correctly; namely, it preserves consistency without deadlock or livelock among operations of the transactions in the system.