L. Bouganim, D. Florescu, and P. Valduriez. Dynamic load balancing in hierarchical parallel database systems. In VLDB, 1996.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....and condition itself are trees, which represent expensive array operations on the multidimensional data. Details on query optimization and execution in array DBMS can be found in [10] 3 Parallel Query Processing Parallel query processing is a well established mechanism in relational DBMS [4] [6] 9] 11] Different hardware architectures have been investigated regarding parallelism, i.e. multiprocessor computer (shared everything, symmetric multiprocessing, SMP) and shared disc shared nothing systems (e.g. workstation cluster) Considering the facts that queries on array data are ....

Bouganim, L., Florescu, D., Valduriez, P.: Dynamic Load Balancing in Hierarchical Parallel Database Systems. In Proc. of the Int. Conf. on Very Large Data Bases (VLDB), Mumbai (Bombay), India, 1996

....strategies for pipelining segments in shared everything environments. Shekita et al. 25] mention synchronous pipelining (SP) a strategy that is again similar to our base model, DTE. But as the paper does not focus on execution, problems like data skew are not considered. Bouganim et al. [7] present dynamic processing (DP) an improved version of PE that uses redistribution. Here, load balancing is achieved by using threads 500 1000 1500 2000 2500 3000 3500 DTE R Figure 10: DTE vs. DTE R (one per processor) that are able to perform each operator. Thus, the PAP can be ....

L. Bouganim, D. Florescu and P. Valduriez. Dynamic Load Balancing in Hierarchical Parallel Database Systems. In Proc. Int'l. Conf. on Very Large Data Bases, Bombay, India, September 1996.

....evaluation strategies for pipelining segments in shared everything environments. Shekita et al. SYT93] mention synchronous pipelining (SP) a strategy that is similar to our base model, DTE. But as the paper does not focus on execution, problems like data skew are not considered. Bouganim et al. BFV96] present dynamic processing (DP) an improved version of PE that uses redistribution. Here, load balancing is achieved by using threads (one per processor) that are able to perform each operator. Thus, the PAP is avoided. DP represents operator parallelism, but the performance of each operator is ....

L. Bouganim, D. Florescu, and P. Valduriez. Dynamic Load Balancing in Hierarchical Parallel Database Systems. To appear in: Proc. Int'l. Conf. on Very Large Data Bases, Bombay, India, September 1996.

....and extend this approach in [10] to scheduling of separate pipelining segments. For their simulation model, they assume uniformly distributed attributes and exclude skew situations. In addition to our effort, Bouganim et al. have independently obtained a partly optimal algorithm, similar to DTE [1]. Road Map. In Section 2, we develop a dataflow model that reflects the relational algebraic structure of the query to be processed. Section 3 describes the processing model to capture the 2. Dataflow Model 3 parallel processing environment. The basic principles of the algorithms and an ....

L. Bouganim, D. Florescu, and P. Valduriez. Dynamic Load Balancing in Hierarchical Parallel Database Systems. Technical Report 2815, INRIA, March 1996.

....through interconnect. The most serious problem for shared nothing systems is load balancing [7] To manage this problem we have to dynamically reorganize the data in the database in proper way. Such reorganization usually requires huge interconnect traffic. An interesting decision is proposed in [8, 9]. It s based on a hierarchical parallel system whose nodes are shared memory multiprocessor. 2 Interprocessor Communication Support in the Omega Parallel Database System In this paper we consider another kind of hierarchical architecture used for parallel DBMS Omega [10] Omega system consists ....

Bouganim L., Florescu D., Valduriez P. "Dynamic Load Balancing in Hierarchical Parallel Database Systems". In: VLDB'96, Proceedings of 22th International Conference on Very Large Data Bases, Mumbai (Bombay), India. September 1996, pp 436-447.

....corresponding consumer, which in the context of databases corresponds to sending several tuples at the same time between two successive operations. It appears that this technical issue has been poorly studied for shared nothing systems. We can nd some bu ered pipelining in hierarchical systems [4], or some tuple by tuple pipelining in shared nothing architectures [9, 14] but no explicit macro pipelining. However, a tuple by tuple inter node pipelining brings very high communication overheads, while a macro pipelining would bring very limited ones, especially 5 1 Referencing Fragment ....

L. Bouganim, D. Florescu, and P. Valduriez. Dynamic load balancing in hierarchical parallel database systems. In Proceedings of the 22 nd VLDB Int'l Conference, pages 436447, Bombay, India, September 1996.

....Parallelism has been recognized as a powerful and cost effective means of handling the projected increases in data size and query complexity in future database applications. Among all proposals, the sharednothing [DG92] and, recently, the more general hierarchical (or, hybrid) NZT96, BFV96] multiprocessor architectures have emerged as the most scalable to support very large database management. In these systems, each site consists of its own set of local resources and communicates with other sites only by message passing (Figure 1) Despite the popularity of these architectures, ....

....strategies for determining the degree of intra operation parallelism and assigning processors in shared nothing DBMSs. Both of these papers avoid dealing with complex query scheduling issues by assuming workloads consisting of simple binary joins and or OLTP transactions. Bouganim et al. BFV96] propose methods for optimizing load balancing on each site of a hierarchical architecture at run time so that inter site data transfers are minimized. In their model, the optimizer still has to determine the assignment of operators to sites and the run time environment has to make up for ....

[Article contains additional citation context not shown here]

Luc Bouganim, Daniela Florescu, and Patrick Valduriez. "Dynamic Load Balancing in Hierarchical Parallel Database Systems". In Proceedings of the 22nd International Conference on Very Large Data Bases, pages 436--447, Mumbai(Bombay), India, September 1996.

....queries and operators. We present heuristic algorithms for various forms of the problem, some of which are provably near optimal. Preliminary experimental results confirm the effectiveness of our approach. 1 Introduction In the shared nothing [7] and the more general hierarchical (or, hybrid) [2] multiprocessor architectures, each site consists of its own set of local resources and communicates with other sites only by message passing. Despite the popularity of these architectures, the development of effective and efficient query processing and optimization techniques to exploit their ....

....parallel query optimization [7] 1 Due to space constraints, we do not discuss the details of earlier work. For an extensive bibliography, the interested reader is referred to the full version of the paper [12] 2 Problem Formulation 2. 1 Definitions We consider hierarchical parallel systems [2] with identical multiprogrammed resource sites connectedby an interconnection network. Each site is a collection of d TS resources (e.g. CPU(s) disk(s) and network interface(s) or communication processor(s) and s SS resources (e.g. memory) Although memory is probably the only SS resource ....

L. Bouganim, D. Florescu, and P. Valduriez. "Dynamic Load Balancing in Hierarchical Parallel Database Systems". In Proc. of the 22nd Intl. VLDB Conf., September 1996.

No context found.

L. Bouganim, D. Florescu, and P. Valduriez. Dynamic load balancing in hierarchical parallel database systems. In VLDB, 1996.

No context found.

Bouganim, L., Florescu, D., Valduriez, P.: Dynamic Load Balancing in Hierarchical Parallel Database Systems. In Proc. of the Int. Conf. on Very Large Data Bases (VLDB), 1996

No context found.

L. Bouganim, D. Florescu, P. Valduriez. Dynamic Load Balancing in Hierarchical Parallel Database Systems. Int. Conf. on VLDB, 1996.

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