O. Ulusoy, G. G. Belford `Real-Time Transaction Scheduling in Database Systems', Information Systems, vol.18, pp.559-580, 1993.  Home/Search   Document Details and Download   Summary   Related Articles   Check

.... may suffer from an unlimited amount of priority inversion time, where priority inversion is a situation in which a higher priority transaction is blocked by a lower priority transaction [28] In the past decade, researchers have proposed various real time concurrency control protocols, e.g. [1, 8, 24, 29, 28, 30, 36], for single site as well as distributed RTDBS. In particular, 8, 28] proposed the idea of priority inheritance, which lets a lower priority transaction inherit the priority of a higher priority transaction which is blocked by the lower priority transaction, to reduce the number of priority ....

O. Ulusoy and G.G. Belford. Real-time Transaction Scheduling in Database Systems. Information Systems, 18(8):559-580 (1993).

....execution histories (section 7.5) In general, first, the size of the transaction is determined by randomly choosing the number of database operations in the transaction from a given range (NumberOps) using a uniform distribution function. These are the similar parameters used in related works [Ulu1 94] A database operation is either a read or a write operation with the probability that WriteProb percent of database operations are write operations. And each database operation is chosen to operate on a database page (disk page) randomly chosen among a given number of database pages ....

....In figures presenting the results in later chapters, performance in terms of success ratios is shown against the mean transaction arrival time which is the mean time between two successive transaction arrivals. This value is exponentially distributed to get fairly independent transaction arrivals [Ulu1 94] In earlier work on RTDBMS, because the flat transaction model is used, another important metric used in evaluating the results is the restart ratio of the system, that is, the number of transactions restarted against the number of transaction arrivals. For flat transactions, restarting a ....

[Article contains additional citation context not shown here]

O. Ulusoy, G.G. Belford, "Real-Time Transaction Scheduling in Database Systems", Information Systems, vol18, no.8, 1994.

....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 ....

....the executing committing conflict an important problem for the performance of a DRTDBS. However, to our best knowledge, the executing committing conflict problem has not received adequate attention in the past. Proposals started to appear in the literature only recently. Although several RT CCPs [5, 7, 13, 14, 15, 16] have been proposed, they are mainly for solving the executing executing conflicts. The previous studies are [1, 2, 17, 18] In [17] and [18] compensation approaches are suggested in which the participants of a transaction are allowed to commit unilaterally, with an attempt to reduce the ....

[Article contains additional citation context not shown here]

Ulusoy, O. and Belford, G. G. (1993) Real-time transaction scheduling in database systems. Inf. Syst., 18, 559--580.

.... [AG89, HCL92, ORP96, YWLS92] In the past two decades, researchers have proposed various efficient real time concurrency control techniques to either reduce the number of deadline violations for soft (and firm 1 ) real time transactions or guarantee the deadlines of hard real time transactions [AG92, BN96, HCL90, HS99, HSR92, KWL99, SLC91, SRSC91, UB93, UB98]. Some of the potential applications of RTDBS are programmed stock trading systems [AGK95, KR99] air traffic management systems, air navigation systems, and critical patient monitoring systems [KS96] An important performance index of these systems is to handle events and requests in a ....

....in a MSRTDBS. Instead, we would like to study how EDF and the priority mapping problem in EDF affects the performance of different strategies in concurrency control. 3 Database System Model The purpose of this section is to define a MSRTDBS model which is based on previous studies in the area [AG92, TSH96, UB93]. It is assumed that the database system consists of three major components [BHG87] the transaction manager (TM) the scheduler, and the data manager (DM) TM maintains a transaction table to record the execution status of transactions in the system and to preprocess transactions if necessary. TM ....

Ulusoy, O. and G.G. Belford, "Real-time Transaction Scheduling in Database Systems", Information Systems, vol. 18, no. 8, pp. 559-580, 1993.

....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 ....

....executing committing conflict an important problem to the performance of a DRTDBS. However, to our best knowledge, the executing committing conflict problem has not been receiving adequate attention in the past. Proposals started to appear in the literatures only recently. Although several RT CCPs [5, 7, 13, 14, 15, 16] have been proposed, they are mainly for solving the executing executing conflicts. The previous studies are [1, 2, 17, 18] In [17] and [18] compensation approaches are suggested in which the participants of a transaction are allowed to commit unilaterally with an attempt to reduce the ....

[Article contains additional citation context not shown here]

Ulusoy, O. & Belford, G.G. (1993) Real-time Transaction Scheduling in Database Systems. Information Systems, Volume 18, Number 8, pp. 559-580.

....in a specified period of time because the price of the stock changes very fast. However, such failures would not be catastrophic. These systems are called soft or firm real time systems. Although significant research efforts have been made in the real time database area in last several years [1, 5, 8, 9, 10, 24, 28, 29], they are mainly focused on scheduling transactions in soft or firm realtime applications. As their applications do not require stringent timing constraints, most of their protocols adopt the best effort approach to scheduling transactions without a guarantee of meeting transaction s deadlines. ....

....transactions at the expense of abortion and re execution overheads, they complicate the system schedulability analysis. Some of them [25] cannot even provide the schedulability analysis since they cannot bound the number of abortions that a lower priority transaction may experience. Ulusoy et al. [29] proposed a data priority based locking protocol (DPL) In DPL, each data item is assigned a priority which is equal to the highest priority of all active transactions that will access the data item. When a transaction comes to the system, the set of data items to be accessed by the transaction is ....

[Article contains additional citation context not shown here]

O. Ulusoy and G.G. Belford. Real-Time Transaction Scheduling in Database Systems. Information Systems, 18(8): 539-557, Elsevier/Pergamon (1993).

....is aptly titled Less Optimism About Optimistic Concurrency Control . OCC implementations tend to be rare and restrictive. Does OCC really help The Optimistic Protocol performed well only under light transaction loads, or when the data resource contention in the system was low according to [Ulusoy 93] Other researchers have made similar observations. The essential difference between conventional 2PL and OCC is that 2PL uses blocking and OCC uses restarts. Belief that OCC is superior to 2PL therefore requires belief that restarts are somehow less harmful than blocking. We find that hard to ....

Ulusoy, Ozgur and Belford, Geneva, "Real-Time Transaction Scheduling in Database Systems," Information Systems, Vol. 18, No. 8, pages 559-580, 1993.

....cause a portion or all of the current version to be lost [138] In the remainder of this section, we discuss the issues related to processing real time transactions. Evaluations of the techniques discussed, if any, are invariably by either a testbed system, e.g. 87] or through simulation, e.g. [3, 76, 215]. One exception is the approximate analysis of real time DBs [79] where an analytical approach is used to approximate the steady state fraction of real time transactions that complete successfully. 6.2.1 Processing Transactions with Hard Deadlines All transactions with hard deadlines must ....

.... for real time databases that use serializability as the correctness criteria include lock based protocols such as two phase locking and its variants [3, 1, 7, 60, 86, 88, 179, 180, 193, 214, 216] optimistic concurrency control protocols [74, 78, 88, 121] and timestamp ordering protocols [133, 195, 215]. Using any of these techniques, conflicts between two real time transactions or between one real time transaction and a set of real time transactions are detected. For lock based protocols, transaction conflicts are resolved by either transaction blocking or transaction abort. When a data item ....

[Article contains additional citation context not shown here]

Ulusoy, O., and Belford, G. G., "Real-Time Transaction Scheduling in Database Systems," Information Systems Journal 18, 8, 1993.

....for this subtransaction is chosen randomly from the given interval NumberOps. Each operation is labeled as a database operation (read write) or as a subtransaction activation point. The probability that an operation is a subtransaction 1 These are the similar parameters used in related works [19]. Nested Transactions in Real Time Database Systems 15 CPU queue cpu2 cpu1 disk1 queue disk2 queue disk2 disk1 cpu n disk n disk n queue Transaction Generator Transaction Queue Transaction Manager Concurrency Control Resource Manager Figure 8 Simulation system components activation point is ....

....before their deadlines. In figures, RT NT performance in terms of success ratios is shown against the mean transaction arrival time which is the mean time between two successive transaction arrivals. This value is exponentially distributed to get fairly independent transaction arrivals [19]. In earlier work on RTDBMS, because the flat transaction model is used, another important metric used in evaluating the results is the restart ratio of the system, that is, the number of transactions restarted against the number of transaction arrivals. For flat transactions, restarting a ....

O. Ulusoy, G.G. Belford, "Real-Time Transaction Scheduling in Database Systems", Information Systems, vol18, no.8, 1994.

....of which transaction to abort and when to resort to priority inheritance must be made judiciously, as in the case of LC, a new protocol. With regard to the last two items, in some sense, the conclusions are not startling. Similar observations have been made in earlier simulation based studies (see [1, 10, 14, 15], for example) or the rarer implementation based study [9] Thus, in terms of performance, the actual performance trends in our results were similar to those in the literature, even though if one were to run our transactions in simulation mode the overheads assumed may affect the actual ....

O. Ulusoy and G.G. Belford, Real-Time Transaction Scheduling in Database Systems, Information Systems, Vol 18, No. 8, pp. 559-580, 1993.

....it is difficult to find a single formula assessing both measures together. Instead, we will observe both metrics to evaluate the system performance. Success ratio, or miss ratio which is the inverse of the success ratio, is the metric used in almost all real time transaction scheduling research [Ulu1 94] Since only flat transactions are considered in RTDBMSs research, no other performance metrics is devised to measure the performance of partially executed transactions. Average subset completion ratio we defined above serves for this purpose. We consider a priority based preemptive scheduling ....

....among transactions. The concurrency control unit manages and coordinates the database accesses through a concurrency control protocol. Currently the system is using the two phase locking (2PL) and timestamp ordering (TO) protocols. 2 These are the similar parameters used in related works [Ulu1 94] After starting execution, transaction first gets CPU, locks the disk page being accessed in its first operation in LockP roc time. Then the transaction gets Disk and reads the disk page in DISKtime time, and gets CPU again for the processing of the page in CPUtime time. Same procedure is ....

O. Ulusoy, G.G. Belford, "Real-Time Transaction Scheduling in Database Systems", Information Systems, vol18, no.8, 1994.

....second. We experimented with other values from one quarter of a second to two seconds, verifying that the relative shape of the results is unaffected by this parameter. We note that the database literature often uses the negative exponential distribution for human think times between transactions [83, 94]. 8 So does the low priority task, but it has no effect, since it has the minimum priority. 116 6.4.2 Statistics and Stopping Criteria The statistics we collected for the simulation runs include the percentage of time there was priority inversion (i.e. the customer holding the lock had ....

....This is insufficient when critical sections may take a long time to execute, such as during periods of scarce resources. Locking protocols in the real time database literature typically assume that tasks (transactions) can be aborted if a higher priority task arrives that conflicts with it [91, 94]. In that domain, all priority inversion can be avoided at the cost of re starting aborted transactions. In our domain, we cannot safely assume that tasks in critical sections can be aborted and restarted. Of the non preemptable locking protocols, the closest to that of priority mediated locks is ....

O. Ulusoy and G. G. Belford. Real-time transaction scheduling in database systems. Information Systems, 18(8):559--80, December 1993.

....of data pages stored in the database. Transaction arrivals are assumed to be exponentially distributed with the mean value of arrival rate. To prevent the possibility of transaction overload in the system, the total number of active transactions in the system is limited by the parameter max mpl [15]. The number of data objects accessed by a transaction is determined by the parameter trans size, which is uniformly distributed between 5 and 20 [2] Page accesses are exponentially distributed across the whole database. For each data access of a normal transaction, the probability that the ....

....parameter trans size, which is uniformly distributed between 5 and 20 [2] Page accesses are exponentially distributed across the whole database. For each data access of a normal transaction, the probability that the accessed data object will be updated is determined by the parameter prob write [15]. The type of the accessed data object is decided based on the parameter per temporal. Temporal data valid intervals are uniformly distributed within the range specified by min interval and max interval. slack factor is used in assigning new transaction deadlines [15] The CPU power, memory ....

[Article contains additional citation context not shown here]

O. Ulusoy, G. Belford, "Real-Time Transaction Scheduling in Database Systems", Information Systems, Vol. 18, No. 8, 1993, pp. 559-580.

No context found.

O. Ulusoy, G. G. Belford `Real-Time Transaction Scheduling in Database Systems', Information Systems, vol.18, pp.559-580, 1993.

....This condition prevents deadlocks if we assume that the real time priority of a transaction does not change during its lifetime and that no two transactions have the same priority. The performance of these two approaches have been studied by some researchers either using simulation (e.g. [2, 4, 26, 65]) or on a RTDBS testbed (e.g. 31, 32] Although, in all those works, both schemes were found to perform better (i.e. satisfy more deadlines) than the conventional 2PL protocol, the results obtained for the comparative performances of the schemes do not completely agree. It was observed by ....

....works, both schemes were found to perform better (i.e. satisfy more deadlines) than the conventional 2PL protocol, the results obtained for the comparative performances of the schemes do not completely agree. It was observed by Huang et al. 31, 32] Haritsa et al. 26] and Ulusoy and Belford [65] that the performance provided by PI cannot reach the level achieved by PA. Remember that PA never blocks high priority transactions, but instead aborts low priority transactions when necessary. It also eliminates the possibility and cost of deadlocks. The authors thus conclude that aborting a low ....

[Article contains additional citation context not shown here]

O. Ulusoy, G. G. Belford `Real-Time Transaction Scheduling in Database Systems', Information Systems, vol.18, no.8, 1993, pp.559-580.

....time critical scheduling methods to observe the timing requirements of transactions [30] A considerable amount of RTDBS research has been devoted to performance evaluation of time cognizant concurrency control protocols. However, the performance studies were either based on simulation (e.g. [1, 2, 3, 10, 11, 20, 23, 27, 28, 29]) or carried out on a RTDBS testbed (e.g. 14, 15] To the best of our knowledge, no analytic performance study has been reported so far involving the evaluation of concurrency control protocols in RTDBSs, which is the main contribution of this paper. 1 The behavior of concurrency control ....

....transaction. HP is characterized by its simplicity and low implementation overhead compared to the other RTDBS concurrency control protocols appeared in the literature. Also, in an earlier simulation work, we found that it can outperform other protocols under a variety of execution environments [29]. Although in this paper we concentrate on the evaluation of protocol HP, the ideas presented are also applicable to the analysis of other concurrency control protocols developed for RTDBSs. In protocol HP, the winner in the case of a lock conflict between two transactions is always the higher ....

[Article contains additional citation context not shown here]

O. Ulusoy, G. G. Belford `Real-Time Transaction Scheduling in Database Systems', Information Systems, vol.18, pp.559-580, 1993.

....simultaneously. There has been a considerable amount of work in the area of RTDBSs. An extensive exploration of the issues is presented in [11, 19, 21] Most of the work in the RTDBS area has focused on the development and evaluation of priority cognizant concurrency control protocols (e.g. [1, 2, 4, 6, 8, 10, 12, 17, 18, 20]) Many proposed concurrency protocols have transferred assumptions and results from non RTDBSs in an unchallenged form to the realm of RTDBSs. The chain of thought underlying conventional concurrency control is the following: I O is 3 4 orders of magnitude slower than a main memory access. ....

....increasing the value of iat (i.e. decreasing the level of transaction load) leads to better performance for all concurrency control protocols tested in this experiment. The comparative performance results of protocols PA, PI, and OPT are similar to what we obtained with a disk resident RTDBS [20]: Protocol PA works in general better than protocol PI for a wide range of mean interarrival time. The optimistic protocol OPT performs well for large values of interarrival time; i.e. when the system is lightly loaded. Since the number of conflicts is small under low load levels, only a few ....

O. Ulusoy and G.G. Belford. Real-time transaction scheduling in database systems. Information Systems, 18(8):559--580 (1993).

....time critical scheduling methods to observe the timing requirements of transactions [17, 30] A considerable amount of RTDBS research has been devoted to performance evaluation of time cognizant concurrency control protocols. However, the performance studies were either based on simulation (e.g. [1, 8, 10, 14, 15, 19, 27, 28]) or carried out on a RTDBS testbed (e.g. 13] To the best of our knowledge, no analytic performance study has been reported so far involving the evaluation of concurrency control protocols in RTDBSs, which is the main contribution of this paper. 1 The behavior of concurrency control ....

....transaction. HP is characterized by its simplicity and low implementation overhead compared to the other RTDBS concurrency control protocols appeared in the literature. Also, in an earlier simulation work, we found that it can outperform other protocols under a variety of execution environments [28]. Although in this paper we concentrate on the evaluation of protocol HP, the ideas presented are also applicable to the analysis of other concurrency control protocols developed for RTDBSs. In protocol HP, the winner in the case of a lock conflict between two transactions is always the higher ....

[Article contains additional citation context not shown here]

O. Ulusoy, G. G. Belford `Real-Time Transaction Scheduling in Database Systems', Information Systems, vol.18, no.8, pp.559-580, 1993.

....namely read write synchronization and write write synchronization, and integrates the solutions to two subproblems considering transaction priorities [Son et al. 1992] In [Kim and Srivastava 1991] new multiversion concurrency control algorithms were proposed to increase concurrency in RTDBSs. In [Ulusoy and Belford 1993], we described several real time concurrency control protocols and reported their relative performance in a single site RTDBS. The remainder of the paper is organized as follows. The next section describes the transaction processing architectures studied. Section 3 provides the structure and ....

O. Ulusoy, G.G. Belford, Real-Time Transaction Scheduling in Database Systems, Information Systems, 18, 559-580.

.... and their performance evaluation can be listed as: Abbott Garcia Molina 1988] Abbott Garcia Molina 1989] Abbott Garcia Molina 1992] Biyabani et al. 1988] Chakravarthy et al. 1994a] Chakravarthy et al. 1994b] Haritsa et al. 1991] Pang et al. 1992] Purimetla et al. 1994] [Ulusoy Belford 1993]. 2.3.2 Concurrency Control Concurrency control in traditional database systems aims to maintain database consistency. Implementation of concurrency control protocols in real time database systems is difficult due to the conflicting requirements of satisfying timing constraints and maintaining ....

.... 1990] Sha et al. 1988] Sha et al. 1990] Sha et al. 1991] Son 1989] Son Chang 1990] Son Lee 1990] Son 1990] Son et al. 1992a] Son et al. 1992b] Son et al. 1992d] Son et al. 1992e] Son Park 1993] Son Kouloumbis 1993] Tseng Chin 1993] Ulusoy Belford 1992] [Ulusoy Belford 1993], Ulusoy Buchmann] Wolfe et al. 1990] 2.3.3 Commitment It is suggested in [Soparkar et al. 1992a] that the unpredictability and the cost of standard commitment protocols make them unsuitable for real time database systems. The authors investigate possible methods to make a commit protocol ....

O. Ulusoy, G. G. Belford `Real-Time Transaction Scheduling in Database Systems ', Information Systems, vol.18, no.8, 1993, pp.559-580.

No context found.

Ulusoy, O. and Belford, G.G., "Real-Time Transaction Scheduling in Database Systems", Information Systems Journal, Vol. 18, No. 8, 1993, pp. 559-580.

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