Abstract. In this paper, we describe the design and implementation of the ADMS query optimizer. This optimizer integrates query matching into optimization and generates more e cient query plans using cached results. It features data caching and pointer caching, alternative cache replacement strategies, and di erent cache update methods. A comprehensive set of experiments were conducted using a benchmark database and synthetic queries. The results showed that pointer caching and dynamic cache update strategies substantially saved query execution time and, thus, increased query throughput under situations with fair query correlation and update load. The requirement of the disk cache space is relatively small, and the extra optimization overhead introduced is more than o set by the time saved in query evaluation. 1

In distributed query processing, the conventional approach to reduce the amount of data transmission is to first apply a sequence of semijoins as "reducers" and then ship the resultant relations to the final site to carry out the join operations. Recently, it has been shown that the approach of applying a combination of joins and semijoins as reducers can lead to substantially larger reduction on data transmission required. In this paper, we develop an efficient heuristic approach to determine an effective sequence of semijoin and join reducers. Semijoins whose execution will reduce the amount of data transmission required to perform a join sequence are termed beneficial semijoins for that join sequence. Note that beneficial semijoins include the conventional profitable semijoins and the gainful semijoins that are not profitable themselves but become beneficial due to the inclusion of join reducers. This type of dependency between semijoin and join reducers complicates the identificati...

In this paper we address the problem of devising a set of indexes for a nested object hierarchy in an object-oriented database to improve the overall system performance. It is noted that the effects of two indexes could be entangled in that the inclusion of one index might affect the benefit achievable by the other index. Such a phenomenon is termed index interaction. Clearly, the effect of index interaction needs to be taken into consideration when a set of indexes is being built. The index selection problem is first formulated and four index selection algorithms are evaluated via simulation. The effects of different objective functions, which guide the search in the index selection algorithms, are also investigated. It is shown by simulation results that the greedy algorithm which is devised in light of the phenomenon of index interaction performs fairly well in most cases. Sensitivity analysis for various database parameters is conducted. Index Term: Object-oriented databases, indexing, nested object hierarchy, index interaction.

Abstract. We present an architecture for query processing in the relational model extended with transaction time. The architecture integrates standard query op-timization and computation techniques with new differential computation tech-niques. Differential computation computes a query incrementally or decremen-tally from the cached and indexed results of previous computations. The use of dif-ferential computation techniques is essential in order to provide efficient process-ing of queries that access very large temporal relations. Alternative query plans are integrated into a state transition network, where the state space includes backlogs of base relations, cached results from previous computations, a cache index, and intermediate results; the transitions include standard relational algebra operators, operators for constructing differential files, operators for differential computation, and combined operators. A rule set is presented to prune away parts of state tran-sition networks that are not promising, and dynamic programming techniques are used to identify the optimal plans from the remaining state transition networks. An extended logical access path serves as a "structuring " index on the cached re-suits and contains, in addition, vital statistics for the query optimization process (including statistics about base relations, backlogs, and queries--previously com-puted and cached, previously computed, or just previously estimated).

The benefit of using indexes for processing queries in a database system is well known. The use of indexes in dismbuted database systems is equally justified. In a distributed database environment a relation may be horizontally partitioned across the nodes of the system and indexes may be created for the fragment of the relation that resides at each node. However, as an alternative. one might wnsuuct each index on the entire relation, i.e., global indexes, and then partition each index between the nodes. Two approaches are presented for processing such an index partitioning scheme in response to a range query and their performance is compared with the typical scheme. The performance of these schemes is evaluated in terms of the response time, system throughput network utilization and disk utilization while varying the number of nodes and query mix.