Next: About this document ...
Up: Related Work [cont'd.]
Previous: Related Work [cont'd.]
- They proposed three algorithms for processing joins in the presence of AVS.
- (i) Tuple interleaving parallel hash join : Based on Bucket spreading hash join discussed above , using a s/ware node coordinator rather than a Omega network.
m (ii)Adaptive Load Balancing parallel hash join : Avoids massive redistribution.In case of mild skew a more selective redistribution is likely to perform better. In this algorithm in the last phase instead of immediately performing local joins a partition tuning phase is done .In this best fit decreasing heuristic is used to determine which bucket to retain locally versus which ones to redistribute.
DBMS
1999-03-11