R* optimizer validation and performance evaluation for local queries

Improving distributed join efficiency with extended bloom filter operations

“…The rest of the records are then sent to site S , where the actual join can occur, and any false positives can be filtered. [18] shows that Bloomjoin consistently outperforms the basic semi-join algorithm.…”

“…The Bloomjoin algorithm [5,18] reduces the amount of data transmitted further by encoding π a (S ) in a bloom filter. It proceeds as follows: First site S produces a bloom filter BF S , including the join key S .a for all its records.…”

Robust Placement of Mobile Relational Operators for Large Scale Distributed Query Optimization

“…Nonetheless, none of the scheduling strategies developed in these fields meet the requirements of the large scale environment where the sources are heterogeneous and data and network characteristics are unpredictable. Simplified methods of scheduling applicable to large scale come from the field of distributed databases exploiting independent parallelism [10,23,26,32,41]. The main idea is to place the relational operators near to data sources or place them on the server site, again based on singlepoint estimations at compile-time.…”

Evaluation of Join Strategies for Distributed Mediation

“…In System R*, both ship-in and ship-out strategies are examined. In [15] a diskbased ship-in strategy (named ship-whole) is implemented with a disk based b-tree index. This type of implementation leads to considerably different results where the ship-out method always outperforms ship-in.…”

“…Disk-based semi-join algorithms are described in [1,2,5,14]. A sort-merge join, bloom filter semi-join, and sort-based semi-join are evaluated in [15] for a distributed database environment. A bloom filter phase can be added to the shipout algorithms described in this paper.…”