. Ulusoy: "Processing Real-Time Transactions in a Replicated Database System", Journal on Distributed and Parallel Databases, Vol. 2, No. 4, pp. 405-436, October 1994.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....deadlines in a DRTDBS. One of the most important factors is data conflicts amongst transactions [4] One kind of conflict occurs among executing transactions, referred to as executing executing conflicts. Most of the proposed real time concurrency control protocols (RT CCPs) for DRTDBSs [5, 6, 7, 8] focus on resolving this kind of conflict. Another kind of conflict involves committing transactions. When a commit protocol works with a concurrency control protocol, data conflicts among executing and committing transactions may occur. Traditionally, they are resolved by blocking, e.g. the ....

....a DRTDBS, the global database is partitioned into a collection of local databases, which are distributed over the sites in the system as shown in Figure 1. Since the performance of a DRTDBS with a sequential transaction model has been found to be similar to a parallel operations transaction model [8], in order not to complicate the transaction model we have chosen the sequential transaction model. In this model, a transaction is defined as a sequence of operations. Each transaction is assigned a deadline based on the application requirements. It is assumed that the transactions are firm ....

[Article contains additional citation context not shown here]

Ulusoy, O. (1994) Processing of real-time transactions in a replicated database system. Distrib. Parallel Databases, 2, 405--436.

.... may occurs [Rajk91] which is highly undesirable to real time scheduling as it can greatly degrade the schedulability of the adopted CPU scheduling algorithm and affect the system performance [Liu73, Stan95] In recent years, different real time concurrency control protocols have been proposed [Agra95, Best95, Hari92, Huan92, Lam95, Lee94, Ulus94]. Most of them are based on 2PL owing to its popularity and simplicity in implementation [Eswa76] The correctness notion adopted in these protocols is serializability [Bern87] For example, in High Priority Two Phase Locking (H2PL) if the requesting lock is being held by a lower priority ....

O. Ulusoy, "Processing Real-Time Transactions in a Replicated Database System", Journal of Distributed and Parallel Database, volume 2, number 4, 1994, pages 405-436.

.... made, they have resulted in schemes wherein either the standard notions of database correctness are not fully supported [19, 20] the maintenance of multiple historical versions of the data is required [18] or the real time transaction semantics and performance metrics pose practical problems [22]. Further, none of these studies have considered the optimistic two phase locking (O2PL) protocol [4] although it is the best performing algorithm in conventional (non real time) replicated database systems [4] In contrast to the above studies, we focus in this paper on the design of one copy ....

....what really matters in determining performance behavior, have changed less than the decrease in absolute values; 3) our objective is to evaluate the relative performance characteristics of the protocols, not their absolute levels. As in several other studies for replicated databases (for example, [2, 22]) here the database size represents only the hot spots , that is, the heavily accessed data of practical applications, and not the entire database. 7 Experiments and Results Using the firm deadline DRTDBS model described in the previous section, we conducted an extensive set of simulation ....

[Article contains additional citation context not shown here]

Ulusoy, O., "Processing Real-Time Transactions in a Replicated Database System," Distributed and Parallel Databases, 2, pp. 405-436, 1994.

....deadlines in a DRTDBS. One of the most important factors is data conflicts amongst transactions [4] One kind of conflicts occurs among executing transactions, referred to as executing executing conflicts. Most of the proposed real time concurrency control protocols (RT CCPs) for DRTDBS [5, 6, 7, 8] focus on resolving this kind of conflicts. Another kind of conflicts involves committing transactions. When a commit protocol works with a concurrency control protocol, data conflicts amongst executing and committing transactions may occur. Traditionally, they are resolved by blocking, e.g. the ....

....In a DRTDBS, the global database is partitioned into a collection of local databases which are distributed over the sites in the system as shown in Figure 1. Since the performance of a DRTDBS with sequential transaction model has been found to be similar to parallel operations transaction model [8], in order not to complicate the transaction model, we have chosen the sequential transaction model. In this model, a transaction is defined as a sequence of operations. Each transaction is assigned a deadline based on the application requirements. It is assumed that the transactions are firm ....

[Article contains additional citation context not shown here]

Ulusoy, O. (1994) Processing of Real-time Transactions in a Replicated Database System. Distributed and Parallel Databases, Volume 2, Number 4, pp. 405-436.

.... intrinsic limitations of mobile computing systems, such as limited bandwidth and frequent disconnection, the design of an efficient and cost effective MDRTDBS requires techniques, which are quite different from those developed for distributed real time database systems (DRTDBS) over wired networks [7,9,10]. How to meet the timing constraints of transactions over a mobile network and, at the same time, to maintain the external and temporal consistency of the databases involves various challenging research issues [10] Up to now, to our knowledge, it is still lack of a detailed study on the design of ....

....of a transaction, the mobile client will generate another transaction after a think time. 4. 2 Model Parameters and Performance Measures Similar to many previous studies on RTDBS and DRTDBS, the deadline of a transaction, T, is defined according to the expected execution time of a transaction [1,9]. The following table lists the model parameters and their base values. Parameters Baseline Values Number of MTSO 1 Number of Cells 7 Location Update Interval 0.2 second Transmission Speed for Channel 10 kbps Number of Channels for each Cell 10 Think Time 8 seconds Transaction Size 7 to 14 ....

O. Ulusoy, "Processing of Real-time Transactions in a Replicated Database Systems", Journal of Distributed and Parallel Databases, vol. 2, no. 4, pp. 405-436 1994.

....Science, Univ. of Massachusetts, Amherst, MA 01003, USA and the Indian Institute of Technology, Bombay. E mail: krithi cs.umass.edu. R. Gupta is with Goldencom Technologies Inc. Fremont, CA 94539, USA. E mail: ramesh goldencom.com. has been considered to some extent (for example, 28] 37] [43], 45] very little work has been done with regard to the equally important issue of ensuring distributed transaction atomicity. We address this lacuna here. A. Commit Protocols Distributed database systems implement a transaction commit protocol to ensure transaction atomicity. Over the last ....

O. Ulusoy, "Processing Real-Time Transactions in a Replicated Database System," Intl. Journal of Distributed and Parallel Databases, vol. 2, no. 4, 1994.

.... they have resulted in schemes wherein either the standard notions of database correctness are not fully supported [17, 18] or the maintenance of multiple historical versions of the data is required [16] or the real time transaction semantics and performance metrics pose practical problems [19]. Further, none of these studies have considered the optimistic two phase locking (O2PL) protocol [3] although it is the best performing algorithm in conventional (non realtime) replicated database systems [3] In contrast to the above studies, we focus in this paper on the design of one copy ....

....in determining performance behavior, have changed a lot less as compared to the decrease in absolute values; 3) Our objective is to evaluate the relative performance characteristics of the protocols, not their absolute levels. As in several other studies for replicated databases (for example, [1, 19]) here the database size represents only the hot spots , that is, the heavily accessed data of practical applications, and not the entire database. Our goal in this experiment was to investigate the performance of the various conflict resolution mechanisms (PA, PI and PA PB) when integrated ....

[Article contains additional citation context not shown here]

Ulusoy, O., "Processing Real-Time Transactions in a Replicated Database System," Distributed and Parallel Databases, 2, pp. 405-436, 1994.

....(or bad) design by tuning the parameters s ur and s qr . Taking the functions loc(n) qr(r) and ur(r) into consideration, we clearly refrain from assuming uniformly distributed access to data objects across the database. This is another major difference to most models proposed in literature [3,10,11,17,28,29,37 42,43]. The probability of propagation now amounts to r k ur. Usually, the time required to process a transaction is mainly determined by the disk service time [25] This in turn is closely related to the number of data objects referenced. Since transactions that access a small number of data items ....

....response time can be minimized by means of full replication. In figure 2 the maximum throughput is shown as a function of n where the coherency index is taking the values 0, 0.1, 0.25, 0.5, 0. 75 and 1 (from top to bottom) Although most evaluations found in literature consider at most 10 sites [11,28,29,37 43] we run the number of sites up to 100 since we agree that future parallel database systems will consist of hundreds of sites [14,32] If the percentage of updates is not negligible (like 10 in figure 2) throughput does not increase linearly with the number of sites due to propagation (when k ....

Ö. Ulusoy: "Processing Real-Time Transactions in a Replicated Database System", Journal on Distributed and Parallel Databases, Vol. 2, No. 4, October 1994, pages 405-436.

.... they have resulted in schemes wherein either the standard notions of database correctness are not fully supported [23, 24] or the maintenance of multiple historical versions of the data is required [22] or the real time transaction semantics and performance metrics pose practical problems [26]. Further, none of these studies have considered the optimistic two phase locking (O2PL) protocol [4] although it is the best performing algorithm in conventional (non realtime) replicated database systems [4] In contrast to the above studies, we focus in this paper on the design of one copy ....

....our first experiment. With these settings, the database is fully replicated and each transaction executes in a sequential fashion. The parameter values for the CPU, disk and message processing times are similar to those in [4] 5 As in several other studies for replicated databases (for example, [1, 26]) here the database size represents only the hot spots , that is, the heavily accessed data of practical applications, and not the entire database. Our goal in this experiment was to investigate the performance of the various conflict resolution mechanisms (PA, PB, PI and PA PB) when integrated ....

[Article contains additional citation context not shown here]

Ulusoy, O., "Processing Real-Time Transactions in a Replicated Database System," Distributed and Parallel Databases, 2, pp. 405-436, 1994.

....[9] It is vital to the correctness of the system to complete the transactions before their deadlines [3] In DRTDBS, the processing of a transaction is much more complex than that in single site RTDBS. It usually has to create a number of sub transactions to access data objects in different sites [7,13]. The system performance is heavily dependent on the local scheduling of the sub transactions in different sites [6,10] In order to maintain the database consistency and to ensure failure atomicity of transactions [2] a real time concurrency control protocol and an atomic commitment protocol ....

....of concurrency control in DRTDBS is that the principles and strategies used in the conventional concurrency control protocols for data scheduling are not compatible with those used in the well known real time CPU scheduling algorithms. Unbound blocking and priority inversion may occur as a result [7,13]. Different methods for conflict resolution are suggested to resolve this problem [9,14] and to make the scheduling of these protocols as consistent as possible with the real time CPU scheduling algorithm adopted. In real time systems, task deadlines can be ensured by the use of a priority ....

Ulusoy, O. "Processing of Real-time Transactions in a Replicated Database Systems", Journal of Distributed and Parallel Databases, vol. 2, no. 4, pp. 405-436 1994.

.... real time database framework incurs the well known additional complexities that are associated with transaction concurrency control and database recovery in distributed database systems [BHG87, OV91] While the issue of distributed realtime concurrency control has been considered to some extent [SRL88, SK92, UB92, Ulu94b], comparatively little work has been done with regard to distributed real time database recovery. We investigate here the performance implications of supporting transaction atomicity in a distributed real time database system. To the best of our knowledge, this is the first quantitative evaluation ....

O. Ulusoy, "Processing Real-Time Transactions in a Replicated Database System", Intl. Journal of Distributed and Parallel Databases, 2(4), 1994.

.... they have resulted in schemes wherein ei1 ther the standard notions of database correctness are not fully supported [22, 23] or the maintenance of multiple historical versions of the data is required [21] or the realtime transaction semantics and performance metrics pose practical problems [25]. Further, none of these studies have considered the optimistic two phase locking (O2PL) protocol [3] although it is the best performing algorithm in conventional (non real time) replicated database systems [3] In contrast to the above studies, we focus in this paper on the design of one copy ....

....that most replicated RTDBS applications will belong to the former category, MIRROR appears to be the best overall choice for implementation in these systems. 4 Related Work Concurrency control algorithms and real time conflict resolution mechanisms for RTDBS have been studied extensively (e.g. [9, 10, 11, 25]) However, concurrency control for replicated DRTDBS has only been studied in [21, 22, 23, 25] An algorithm for maintaining consistency and improving the performance of replicated DRTDBS is proposed in [21] In this algorithm, a multiversion technique is used to increase the degree of ....

[Article contains additional citation context not shown here]

Ulusoy, O., "Processing Real-Time Transactions in a Replicated Database System," Distributed and Parallel Databases, 2, pp. 405-436, 1994.

....eventually receive all updates [19] Multiple copy updates lead to a considerable overhead due to the communication required among the data sites holding the copies. 18 The impact of storing multiple copies of data on satisfying timing constraints of RTDBS transactions was investigated by Ulusoy [66]. A detailed performance model of a distributed RTDBS is employed in that work to evaluate the effects of various workload parameters and design alternatives on the system performance. The primary performance issue considered is the satisfaction of transaction deadlines; more specifically, an ....

....executed. Distributed versions of the time cognizant concurrency control protocols (see Section 2.4) need to be executed in distributed RTDBS environments. The performance of distributed lock based concurrency control protocols was studied by Ulusoy both in a nonreplicated [64] and a replicated [66] RTDBS environments. The distributed version of the priority abort protocol PA was observed to perform better than the distributed priority inheritance protocol PI under various conditions in a nonreplicated RTDBS [64] However, the difference between the performance results of protocols is not as ....

[Article contains additional citation context not shown here]

O. Ulusoy `Processing Real-Time Transactions in a Replicated Database System', International Journal of Distributed and Parallel Databases, vol.2, no.4, 1994.

....among the data sites holding the copies. The following papers discuss the impact of storing multiple copies of data on satisfying timing constraints of transactions, and propose techniques to enhance the availability of replicated realtime databases: Lin Lin 1988] Son Kouloumbis 1992] [Ulusoy 1994], Ulusoy 1995c] 2.7 Security Security in real time database systems is difficult to achieve due to the conflicting goals of the timing requirement and the security requirement. The security problem in real time databases is addressed in: David et al. 1994] Son Thuraisingham 1993] 2.8 ....

O. Ulusoy `Processing Real-Time Transactions in a Replicated Database System', Distributed and Parallel Databases, vol.2, no.4, 1994, pp.405-436.

No context found.

. Ulusoy: "Processing Real-Time Transactions in a Replicated Database System", Journal on Distributed and Parallel Databases, Vol. 2, No. 4, pp. 405-436, October 1994.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC