Results 1 
4 of
4
Dynamically Distributed Query Evaluation
 In PODS
, 2001
"... Distributed query evaluation usually assumes a fixed topology, where the set of servers and the partitioning of data on the servers is known in advance. Given a query expression, an optimizer will first produce a global plan, then assign ..."
Abstract

Cited by 31 (2 self)
 Add to MetaCart
Distributed query evaluation usually assumes a fixed topology, where the set of servers and the partitioning of data on the servers is known in advance. Given a query expression, an optimizer will first produce a global plan, then assign
Exploring Load Balancing in Parallel Processing of Recursive Queries
, 1996
"... Recent work on load balancing has confirmed its importance when one wants to achieve good performances during the actual evaluation of parallel database queries. Existing work mostly focuses on the join processing for parallel relational databases. We are interested here in more complex queries, suc ..."
Abstract

Cited by 6 (5 self)
 Add to MetaCart
Recent work on load balancing has confirmed its importance when one wants to achieve good performances during the actual evaluation of parallel database queries. Existing work mostly focuses on the join processing for parallel relational databases. We are interested here in more complex queries, such as recursive ones. The main difference is that, in the latter case, the work due to a task cannot be previously determined and, consequently, no method can define at the outset the tasks to be executed in parallel in order to balance the workload at each processor. We propose a dosedriven dynamic strategy that aims at obtaining an improved workload balance and better use of the available resources. We examine the applicability of our strategy with its specialization to the case of the transitive closure query. Preliminary computational results on randomly generated test problems illustrate the efficiency of the proposed method. Keywords: Load Balancing, Dynamic Assignment, Recursive Quer...
General Terms: Theory
"... The degree sequence of an nvertex graph is d0,..., dn−1, where each di is the number of vertices of degree i in the graph. A random graph with degree sequence d0,..., dn−1 is a randomly selected member of the set of graphs on {1,..., n} with that degree sequence, all choices being equally likely. L ..."
Abstract
 Add to MetaCart
The degree sequence of an nvertex graph is d0,..., dn−1, where each di is the number of vertices of degree i in the graph. A random graph with degree sequence d0,..., dn−1 is a randomly selected member of the set of graphs on {1,..., n} with that degree sequence, all choices being equally likely. Let λ0, λ1,... be a sequence of nonnegative reals summing to 1. A class of finite graphs has degree sequences approximated by λ0, λ1,... if, for every i and n, the members of the class of size n have λin + o(n) vertices of degree i. Our main result is a convergence law for random graphs with degree sequences approximated by some sequence λ0, λ1,....With certain conditions on the sequence λ0, λ1,..., the probability of any firstorder sentence on random graphs of size n converges to a limit as n grows.
A Centralized Communication Approach for the Parallel SharedNothing Evaluation of Datalog
"... Communication cost is a fundamental issue for the parallel evaluation of Datalog queries based on a sharednothing architecture. If it is efficiently handled, an optimization of the overall processing time may be achieved. Typically, the evaluation is parallelized by partitioning the rule instantiat ..."
Abstract
 Add to MetaCart
Communication cost is a fundamental issue for the parallel evaluation of Datalog queries based on a sharednothing architecture. If it is efficiently handled, an optimization of the overall processing time may be achieved. Typically, the evaluation is parallelized by partitioning the rule instantiations among a set of processing nodes, with a completely distributed strategy. In this paper we propose a strategy where one coordinating processing node is in charge of the interprocessor data transfers. In many cases, our approach may have better performance compared to the existing strategies. The data transmission criterion defined over such a scheme is more practical, permitting to efficiently determine the transmission sets and to obtain a reduction on the total number of communications and the amount of data transferred among the processing nodes. We give some experimental results that back up a preliminary performance comparison of our approach with respect to previous ones. We argue...