• Journal of Computing Science and Engineering
• /
• v.3 no.1
• /
• pp.27-58
• /
• 2009

Aggregation join queries are an important class of queries over data streams. These queries involve both join and aggregation operations, with window-based joins followed by an aggregation on the join output. All existing research address join query optimization and aggregation query optimization as separate problems. We observe that, by putting them within the same scope of query optimization, more efficient query execution plans are possible through more versatile query transformations. The enabling idea is to perform aggregation before join so that the join execution time may be reduced. There has been some research done on such query transformations in relational databases, but none has been done in data streams. Doing it in data streams brings new challenges due to the incremental and continuous arrival of tuples. These challenges are addressed in this paper. Specifically, we first present a query processing model geared to facilitate query transformations and propose a query transformation rule specialized to work with streams. The rule is simple and yet covers all possible cases of transformation. Then we present a generic query processing algorithm that works with all alternative query execution plans possible with the transformation, and develop the cost formulas of the query execution plans. Based on the processing algorithm, we validate the rule theoretically by proving the equivalence of query execution plans. Finally, through extensive experiments, we validate the cost formulas and study the performances of alternative query execution plans.

• The KIPS Transactions:PartD
• /
• v.16D no.3
• /
• pp.295-306
• /
• 2009

Query execution time minimization is an important objective in distributed database design. While total time minimization is an objective for On Line Transaction Processing (OLTP), response time minimization is for Decision Support queries. We formulate the sub-query allocation problem using analytical models and solve with genetic algorithm (GA). We show that query execution plans with total time minimization objective are inefficient from response time perspective and vice versa. The procedure is tested with simulation experiments for queries of up to 20 joins. Comparison with exhaustive enumeration indicates that GA produced optimal solutions in all cases in much less time.

• Journal of Communications and Networks
• /
• v.16 no.4
• /
• pp.430-435
• /
• 2014

The performance of a distributed database system depends particularly on the site-allocation of the fragments. Queries access different fragments among the sites, and an originating site exists for each query. A data allocation algorithm should distribute the fragments to minimize the transfer and settlement costs of executing the query plans. The primary cost for a data allocation algorithm is the cost of the data transmission across the network. The data allocation problem in a distributed database is NP-complete, and scalable evolutionary algorithms were developed to minimize the execution costs of the query plans. In this paper, quadratic assignment problem heuristics were designed and implemented for the data allocation problem. The proposed algorithms find near-optimal solutions for the data allocation problem. In addition to the fast ant colony, robust tabu search, and genetic algorithm solutions to this problem, we propose a fast and scalable hybrid genetic multi-start tabu search algorithm that outperforms the other well-known heuristics in terms of execution time and solution quality.

• Journal of KIISE:Computing Practices and Letters
• /
• v.5 no.5
• /
• pp.543-552
• /
• 1999

The iterator technique which is used for implementing physical operators of the query executor is known for its efficiency and extensibility. The most widely used technique for processing an operator on spatial objects is to process by dividing it into the filter step and the refinement step. Recently, there was a research for an optimizer which can generate more efficient query execution plans than those of traditional methods by separating a spatial operator into filter and refinement steps in the level of the object algebra. But, traditional query executors were not designed considering such query execution plans. So they have no function of transmitting the result of the filter operation between operators. We propose two methods, the probe technique and the tag technique, which transmit the result of the filter operator when using the iterator in the query execution plan in which operators are separated by filter/refinement steps and other operators can be allowed between the steps. Whereas the probe technique extends the state record within an operator, the tag technique stores the result of a filter step in an intermediate result in the form of the tag. Based on the comparison of these methods, we design and implement a query executor using the tag technique that is superior in extensibility. The implemented query executor can execute operations defined in the Spatial Object Algebra(SOA) to process an extended OQL for spatial queries.