Mobile Join Algorithms based on Mobiles Agents for Large Scale Distributed Query Optimization

Abstract

In the large scale distributed environment, the query optimization presents new problems because of the data unavailability, the estimations inaccuracies and environment instability. In this paper, we address the sub-optimality of executions plans caused by these problems. We propose to extend the join algorithms based on mobile agents in order to correct the sub-optimality. This extension allows the join to change their execution site. Indeed, the mobile agent executing a join adapts to changes in characteristics of the execution environment (e.g. network bandwidth, available memory) and responds to the estimations inaccuracies (e.g. size of intermediate relations). The performance evaluation shows that the proposed algorithms improve the response time whatever the variation of estimations errors.

Decentralized Dynamic Query Optimization based on Mobiles Agents for Large Scale Data Integration Systems

“…A characteristic of dynamic optimization methods is the decision making nature of the modifications of the execution plans (decision making column in the Table 1). We distinguish two classes: centralized methods [1,2,3,4,5,12,24] and decentralized dynamic optimization methods [9,19,20,21,23,27] A method of optimization known as centralized if it is based on a main process which is charged to supervise, control and modify the execution plans. This process can be based on other modules ensuring the production of necessary information for the control and the modifications of an execution plan.…”

“…During the query optimization, the optimizer of mediator generates an execution plan including : (i) the number of mobile agents participating in the evaluation of execution plan, (ii) the operator will be executed by each agent, and (iii) the execution site of each agent. In the rest of this section, we present the mobile join algorithms [20]. Then, we propose an execution model based on mobile agents for queries including cooperation policy allowing the agents to run efficiently the execution plans.…”