L. Amsaleg, M. Franklin, and A. Tomasic. Dynamic query operator scheduling for wide-area remote access. Journal of Distributed and Parallel Databases, 6(3), July 1998.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....are often affected by the time of day, day, etc. and points of congestion between the client and remote server [21] There has been extensive prior research on query evaluation techniques to accommodate transient behavior. Reactive query evaluation techniques at the plan level are described in [1, 4, 18, 25]. Alternately, adaptive evaluation techniques at the plan and operator implementation level are described in [3, 11, 12, 14, 24] Research in [2, 7, 13, 15, 29] have addressed various aspects of the task of modifying an optimizer to handle distributions for various parameters, e.g. available ....

....on pipelined fragments of query execution. The Telegraph project explores adaptive fine grained data flow based query processing techniques using rivers, eddies, ripple join, XJoin, etc. 3, 11, 24] They also focus on continuous and high frequency response to changing feedback. Query scrambling [1, 25] is a query optimization and evaluation technique to deal with transient conditions, e.g. unexpected delay. It is based on plan re organization to avoid idle time, and the synthesis of new operators to execute in the absence of other work. It makes a key contribution of a response time based ....

L. Amsaleg, M. Franklin, A. Tomasic, and T. Urhan. Dynamic query operator scheduling for wide-area remote access. Journal of Distributed and Parallel Databases, 6(3), 1998.

....1 58113 624 2 03 03 . 5.00. can make any static query plan sub optimal over time. Adaptive query processing has been recognized as an important problem with significant research activity (see Section 6) Most of the previously proposed approaches to adaptive query processing can be classified [1] as either proactive (i.e. calculate the adaptation alternatives at compile time) or reactive (i.e. find appropriate adaptations at run time according to environmental changes) In our adaptive query processing engine, called Ginga, we combine proactive alternative query plan generation with ....

....the query execution to memory and CPU constraints is part of our ongoing research study. 6. RELATED WORK Adapting the execution of queries due to changes to the runtime environment parameters has been an important area of research, starting from early 90 s [6, 5] and continuing to today [2, 1, 7, 8, 3, 4]. Depending on when the adaptation takes place, previous approaches to query adaptation can be broadly categorized into proactive and reactive methods [1] Proactive methods adapt query execution at start up time (i.e. while loading the query plan to be executed) based on the current runtime ....

[Article contains additional citation context not shown here]

L. Amsaleg, M. J. Franklin, and A. Tomasic. Dynamic query operator scheduling for wide-area remote access. Journal of Distr. and Parallel Databases, 6(3), 1998.

.... irt user in intra U O O Mid Query Re optimisation [15] rem any maybe maybe ac statstrt stats inter U L A Eddies [4, 5] op or no no no any trt stats intra U L A Rivers [3] no parl no yes per fl trt l dr intra P O A MIND [18] op or any no yes ac statstrt stats inter D G A Query Scrambling [1, 2, 26] rem any no no dar trt in av inter D G A Pipeline Scheduler [25] op or PJ no no user pr irt dr user in intra D O A Bouganim et al. [6, 7] rem PJ maybe maybe dar mem fl trt dr mem av intra D O G A Conquest [16, 17] rem any yes yes any trt stats inter P G A Tukwila [13, 14] rem ....

....scheduler is invoked less often in order to avoid large overheads. 4.2 Algorithms That Adapt to Data Arrival Rates In the light of data arrival delays, a common approach is to minimise idle time by performing other useful operations, thus attenuating the e#ect of such delays. Query Scrambling [1, 2, 26] and a generic AQP architecture discussed in [6, 7] are two representative examples in this area. Query scrambling focuses on problems incurred by delays in receiving the first tuples from a remote data source. The system performs other useful work in the hope that the problem will eventually be ....

L. Amsaleg, M. Franklin, and A. Tomasic. Dynamic query operator scheduling for wide-area remote access. Distributed and Parallel Databases, 6(3):217--246, 1998.

....deploy such systems, Geographical Information Systems for instance [90] For a survey on Information Integration systems see [36] Very little attention is paid in the development of such systems to optimizing querying times and latencies. At most attempts are made to ensure that delay is hidden [8]. In our view this area is ripe for intervention from experts in data structures, scheduling and online algorithms. Sophisticated Information Integration systems that incorporate the varying latencies of different databases, the transmission times of di erent sized query returns, and other such ....

L. Amsaleg, M. Franklin, and A. Tomasic. Dynamic query operator scheduling for wide-area remote access. Distributed and Parallel Databases, 6(3):217-246, July 1998.

....a desired performance guarantee. 2 There has been extensive prior research on query optimization and evaluation that is relevant to our task. A recent issues of the IEEE Data Engineering bulletin highlights several projects. Reactive query evaluation techniques at the plan level are described in [1, 4, 21, 26]. Alternately, adaptive evaluation techniques at the plan and operator implementation level are described in [3, 13, 14, 16, 25] Research in [2, 7, 19, 15, 17, 32] have addressed various aspects of the task of modifying an optimizer to handle distributions for various parameters, e.g. available ....

....on pipelined fragments of query execution. The Telegraph project explores adaptive fine grained data flow based query processing techniques using rivers, eddies, ripple join, XJoin, etc. 3, 13, 25] They also focus on continuous and high frequency response to changing feedback. Query scrambling [1, 26] is a query optimization and evaluation technique to deal with transient conditions, e.g. unexpected delay. It is based on plan re organization to avoid idle time, and the synthesis of new operators to execute in the absence of other work. It makes a key contribution of a response time based ....

[Article contains additional citation context not shown here]

L. Amsaleg, M. Franklin, A. Tomasic, and T. Urhan. Dynamic query operator scheduling for wide-area remote access. Journal of Distributed and Parallel Databases, 6(3), 1998.

....selectivities, available memory) are far from the estimates made during planning, thus invalidating the QEP [8,9] at the scheduling level, when the query engine faces unpredictable delays while accessing remote data. The query engine then stalls, thereby increasing the response time [1,2,6,15]. at the physical operator level, discovering, for instance, that the available memory for the operator execution is not sufficient [4,8] Poor performance of integration query processing lies in the fact that the execution is fully specified before it starts, and is never revised until it ....

.... three different levels: at the QEP level, by partially re optimizing the query plan in order to adapt to the actual values of cardinality, selectivity and available memory [8,9] at the scheduling level, by modifying on the fly the scheduling of the operators to avoid query engine stalling [1,2,6,15]. at the operator level, using auto adaptive relational operators [4, 8] These techniques are complementary and should be used together to provide good performance [8,15] Indeed, partial re optimization of the query plan is difficult to implement and tune [15] Moreover, the possibility for ....

L. Amsaleg, M. J. Franklin, and A. Tomasic. Dynamic Query Operator Scheduling for Wide-Area Remote Access. Journal of Distributed and Parallel Databases, 6 (3), July 1998. 47

....are caused by the autonomy of the data sources and the unpredictability of the performance of the network. In particular, when accessing web sources we may experience initial delays before data is transmitted, and even when it is, the arrival of the data may be bursty. The work described in [AFT98, UFA98] has considered the problem of adapting a query execution plans to initial delays in the arrival of the data. Wrapper construction: Recall that the role of a wrapper is to extract the data out of a web site into a form that can be manipulated by the data integration system. For example, ....

Laurent Amsaleg, Michael Franklin, and Anthony Tomasic. Dynamic query operator scheduling for wide-area remote access. Distributed and Parallel Databases, 6(3):217--246, July 1998.

....Internet, it is important to assess the impact of the unpredictable network delays that are often found in widearea networks, on the three cache consistency algorithms. Initial message delay, slow delivery and bursty arrival are the three types of delays examined in a recent WAN performance study [AFT97]. Similar to that study [AFT97] we simulate network delay by making the message sending source wait for a specified time before sending the message. The message sending source flips a coin to determine whether a message should be delayed (delay probability) The actual value of the delay (delay ....

....the impact of the unpredictable network delays that are often found in widearea networks, on the three cache consistency algorithms. Initial message delay, slow delivery and bursty arrival are the three types of delays examined in a recent WAN performance study [AFT97] Similar to that study [AFT97], we simulate network delay by making the message sending source wait for a specified time before sending the message. The message sending source flips a coin to determine whether a message should be delayed (delay probability) The actual value of the delay (delay time) is chosen as a multiple ....

L. Amsaleg, M. Franklin, and A. Tomasic. Dynamic Query Operator Scheduling for Wide-Area Remote Access. Technical Report CS-TR-381, University of Maryland Tech Report, 1997.

....to evaluate the assignment of WSIs in the pre plan are useful (as shown in our experiments) there are limitations to this approach. First, several operators and techniques for adaptive query optimization, tailored to overcome delays associated with accessing remote WebSources, have been developed [2, 3, 10, 18, 22, 23, 26, 25, 46, 43]. Typically, these adaptive operators are designed to overcome delays, but obtaining cost formulas for these operators is non trivial. The use of these operators in the actual evaluation plans should improve query execution time. In future work, we will examine the impact of delays and adaptive ....

....[1] and the WebPT [52] is appropriate to modeling the costs of accessing WebSources, since their models use query feedback. Finally, several operators and techniques for adaptive query optimization, tailored to overcome delays associated with accessing remote WebSources, have been developed [2, 3, 10, 22, 23, 26, 25, 43, 46]. While our cost based heuristics have not extensively considered such adaptive techniques, it is clear that adaptive operators will have an impact. This is an area for future research. To conclude, we presented a two phase optimization approach in the WQO.We presented several cost based ....

L. Amsaleg, M. Franklin, A. Tomasic, and T. Urhan. Dynamic query operator scheduling for wide-area remote access. Journal of Distributed and Parallel Databases, 6(3), 1998.

....modify the query in some way so that the answer can be provided with the data that has arrived. One example of such an approach is called Parachute Queries [2, 7] Such semantic approaches, however, are beyond the scope of our current work. 4. 1 Query Scrambling We have developed Query Scrambling [4, 3, 37] to address the issue of unpredictable delays in the wide area environment. Query Scrambling reacts to unexpected delays by modifying, on the fly, the execution plan of a query so that progress can be made on other parts of the plan. In other words, rather than simply stalling for delayed data to ....

Laurent Amsaleg, Michael J. Franklin, and Anthony Tomasic. Dynamic query operator scheduling for widearea remote access. Journal of Distributed and Parallel Databases, 6(3), 1998.

....arrivals. With traditional query processing techniques, query execution can become blocked even if only one of the accessed data sources experiences such delays. We developed Query Scrambling to address this problem and showed how it can be used to hide initial delays [UFA98] and bursty arrivals [AFT98] Query Scrambling is a reactive approach to query execution; it reacts to data delivery problems by on the fly rescheduling of query operators and restructuring of the query execution plan. Query Scrambling is aimed at improving the response time for the entire query, and may actually slow down ....

....other research due to (among other reasons) the focus on coping with unpredictable delays arising from widearea remote data access. As mentioned previously, the XJoin approach is complementary to our earlier work on Query Scrambling for unpredictable delays in distributed query processing [AFTU96, AFT98, UFA98] as well as to other dynamic approaches such as [ONK 97, TRV96] A recent paper [IFFL 99] describes the Tukwilla system that contains an operator similar to XJoin, which can adapt to limited memory, but that differs from XJoin in several ways. Most importantly that operator does ....

L. Amsaleg, M. J. .Franklin, and A. Tomasic. Dynamic Query Operator Scheduling for Wide-Area Remote Access. Journal of Distributed and Parallel Databases, Vol. 6, No. 3, July 1998.

....the DBMS can cope with delays that arise at run time, regardless of whether or not parallelism is pre compiled into a plan. Adapting to unexpected delays requires more flexible approaches, such as those we describe in the following sections. 3 Query Scrambling Using Query Scrambling [AFTU96, AFT97] a query is initially executed according to the original plan and associated schedule generated by the query optimizer. If, however, a significant performance problem arises during the execution, then Scrambling is invoked to modify the execution on the fly, so that progress can be made on other ....

....which can reduce the efficiency of algorithms that could otherwise exploit (faster) sequential I Os. Two factors are relevant here: the speed of the network compared to the speed of the local disk(s) and the way data is obtained through the network (i.e. page at a time versus streaming) AFT97] A second set of tradeoffs revolve around the question of whether to Reschedule individual operators (in a bottom up fashion) or entire subtrees. Subtrees can be rescheduled (i.e. executed out of order) given sufficient available memory. Such rescheduling enables the entire subtree to be ....

L. Amsaleg, M. Franklin, and A. Tomasic. Dynamic query operator scheduling for wide-area remote access. Tech. Report CS-TR-3811 and UMIACS-TR-97-54, Univ. of MD, College Park, July 1997.

....between conservative approaches, which are safer for short delays, and more aggressive approaches which lead to bigger savings in the event of long delays. In general, the amount of delay that can be hidden by scrambling (in the absence of creating additional parallelism, as is discussed in [AFT98] is limited by the normal response time of the query. This is becausescrambling hides delays by performing other useful work, so its ability to hide delay is limited by the amount of useful work that can be done. Thus, as the delay increases beyond the normal responsetime of the original query, ....

....work that can be done on scrambling and related dynamic techniques. First, although not discussed in this paper, the scrambling techniqueswe have described here can be adapted for use with other types of delay, such as bursty arrival, in which sites repeatedly stall and recover. As described in [AFT98] such delays introduce a number of scheduling and memory management issues that must be addressed by scrambling. In addition, we would like to investigate the use of delay prediction techniques in the scrambling approaches. Finally, as described in [ABF 97] additional techniques are required ....

L. Amsaleg, M. J. .Franklin, and A. Tomasic. Dynamic Query Operator Scheduling for Wide-Area Remote Access. Journal of Distributed and Parallel Databases, Vol. 6, No. 3, July 1998.

No context found.

L. Amsaleg, M. Franklin, and A. Tomasic. Dynamic query operator scheduling for wide-area remote access. Journal of Distributed and Parallel Databases, 6(3), July 1998.

No context found.

Amsaleg, L., Franklin, M. and Tomasic, A. "Dynamic query operator scheduling for wide-area remote access," Distributed and Parallel Databases, vol. 6(3): pp. 217-246, 1998.

No context found.

L. Amsaleg, M. Franklin, and A. Tomasic. Dynamic query operator scheduling for wide-area remote access. Distributed and Parallel Databases, 6(3):217--246, July 1998.

No context found.

Amsaleg, L., M. Franklin, and A. Tomasic, "Dynamic query operator scheduling for wide-area remote access," Distributed and Parallel Databases, vol. 6(3): pp. 217-246, 1998.

No context found.

L. Amsaleg, M. Franklin, and A. Tomasic. Dynamic query operator scheduling for wide-area remote access. Journal of Distributed and Parallel Databases, 6(3), July 1998.

No context found.

L. Amsaleg, M. Franklin, and A. Tomasic. Dynamic query operator scheduling for wide-area remote access. Distributed and Parallel Databases, 6(3):217--246, 1998.

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