K. Ono and G. Lohman. Extensible Enumeration of Feasible Joins for Relational Query Optimization. Technical Report RJ 6625 (63936), IBM Almaden Research Center, December 1988.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....the standard for relational low level query interfaces. Similar low level interfaces have not been defined for other data models because they often are mapped to the relational model. Enumeration of alternate operation sequences has received significant attention in the context of join ordering [OL88, RR82, SAC 79] Almost all of the referenced work is applied in this thesis. The proposed query processing methodology, with the exception of the type consistency step, is similar to that proposed by Jarke [JK84] The notion of domain independence in the relational model is extended to define ....

....and infinite classes in Chapter 3 and their relationship to the safety of queries. We use logical, equivalence preserving transformations to improve query expressions and define them with conditions which would be suitable for a rule based transformation system. The join enumeration algorithm of [OL88] and join template representation of [RR82] are modified and extended for generating our access plans. 1.3.4 Query Languages Many query languages have been developed for databases. I make the distinction between user and formal languages. Languages such as QUEL [Sto76] and SQL [Dat87] are ....

[Article contains additional citation context not shown here]

K. Ono and G. Lohman. Extensible Enumeration of Feasible Joins for Relational Query Optimization. Technical Report RJ 6625 (63936), IBM Almaden Research Center, December 1988.

....with a specific number of join operators. Thus, although some efficiency is lost due to the abstractions of OPT , a specific implementation may still over ride the abstractions and achieve efficiency (at the cost of extensibility) In fact, the various join enumeration algorithms described in [OL90] can each be implemented in OPT as a class derived from the TREETOTREEGENERATOR class. 2.5.2 Modifying the Optimizer Changing the logical or physical Algebra : To modify the optimizer to incorporate a new physical implementation algorithm, a new class corresponding to that algorithm must be ....

....with access to the internal data structures of the System R dynamic programming style search strategy that was used) was also coded to give better performance. This resulted in the SMARTBUSHYJOINENUMERATOR and SMARTCARTESIANJOINENUMERATOR classes which are based on the schemes described in [OL90] As an experimental evaluation of the optimizer, we studied its performance (optimization time and estimated execution cost) as a function of the number of joins in the input query. For each query size (number of joins) 10 different queries were generated randomly and optimized. The ....

[Article contains additional citation context not shown here]

K. Ono and G.M. Lohmann. "Extensible Enumeration of Feasible Joins for Relational Query Optimization ". In Proc. of the 16th VLDB Conf., August 1990.

....a pass through the algorithm. Pass 1: Recall that processing template stream nodes represent oid streams which can be combined to evaluate atoms or to remove duplicates. The first step of each pass then, is to enumerate all possible ways of combining stream nodes. We use the algorithm given in [OL88] for join enumeration but modify it slightly such that it does not produce combinations where a stream node is combined with itself (self join) The final node is not included in the enumeration. Enumeration of the initial processing template results in the following permutations of stream nodes. ....

....by size. 1: 1 2 3 4 5 2: 1,2 1,3 2,3 1,4 2,4 3,4 1,5 2,5 3,5 4,5 3: 1,2,3 1,2,4 1,3,4 2,3,4 1,2,5 1,3,5 2,3,5 1,4,5 2,4,5 3,4,5 4: 1,2,3,4 1,2,3,5 1,2,4,5 1,3,4,5 2,3,4,5 5: 1,2,3,4,5 Similar to the filtering process described in [OL88] each permutation is tested to determine whether it is a useful combination of stream nodes. Each permutation of stream nodes defines mappings to sets of variables and sets of consumed atoms. For example the permutation f1,2,5g defines the mapping shown in Figure 8. We define two interesting ....

K. Ono and G. Lohman. Extensible enumeration of feasible joins for relational query optimization. Technical Report RJ 6625 (63936), IBM Almaden Research Center, December 1988.

....a pass through the algorithm. 5.1 Pass 1 Recall that processing template stream nodes represent oid streams which can be combined to evaluate atoms or to remove duplicates. The first step of each pass then, is to enumerate all possible ways of combining stream nodes. We use the algorithm given in [22] for join enumeration but modify it slightly such that it does not produce combinations where a stream node is combined with itself (self join) The final node is not included in the enumeration. Enumeration of the initial processing template results in the following permutations of stream nodes. ....

....by size. 1: 1 2 3 4 5 2: 1,2 1,3 2,3 1,4 2,4 3,4 1,5 2,5 3,5 4,5 3: 1,2,3 1,2,4 1,3,4 2,3,4 1,2,5 1,3,5 2,3,5 1,4,5 2,4,5 3,4,5 4: 1,2,3,4 1,2,3,5 1,2,4,5 1,3,4,5 2,3,4,5 5: 1,2,3,4,5 Similar to the filtering process described in [22], each permutation is tested to determine whether it is a useful combination of stream nodes. Each permutation of stream nodes defines mappings to sets of variables and sets of consumed atoms. For example the permutation f1,2,5g defines the mapping shown in Figure 11. We define two interesting ....

K. Ono and G. Lohman. Extensible enumeration of feasible joins for relational query optimization. Technical Report RJ 6625 (63936), IBM Almaden Research Center, December 1988.

....in which the inner relation is a base relation restricts the search space to the space of left deep join trees. On the other hand, implementing a BushyJoinExpand class that considers composite inners will generate all bushy trees. In fact, the various join enumeration algorithms described in [OL90] can each be implemented in OPT as a class derived from the TreeToTreeGenerator class. 3.5.2 Modifying the Optimizer Changing the logical or physical Algebra : To modify the optimizer to incorporate a new physical implementation algorithm, a new class corresponding to that algorithm must be ....

....search spaces on the time taken for optimization, and Figure 18 shows the effect on the relative estimated execution costs of the optimal plans produced. Note that optimization times are shown on a logarithmic scale. 5 These join enumerator classes are based on the schemes described in [OL90] 0 5 10 15 Number of Joins 0.01 0.1 1 10 100 Optimization Time (seconds, logscale) Left Deep Bushy Bushy Cartesian 0 2 4 6 8 10 Number of Joins 1 2 3 Estimated Execution Costs (Scaled) Left Deep Bushy Bushy Cartesian Figure 17: Comparison of Search Spaces: Optimization Times (Log scale) ....

K. Ono and G.M. Lohmann. "Extensible Enumeration of Feasible Joins for Relational Query Optimization ". In Proc. of the 16th VLDB Conf., August 1990.

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