Publication at Faculty of Mathematics and Physics |
2005
Abstract
The DataPile oriented systems ideas were published together with basic replication algorithms. The implementation of such system in a real production environment showed that proposed algorithms are not sufficient, from both point of view of expression power as well as performance.
The system implemented exactly according to the theoretical proposal would not meet user requirements to a replication logic and would have inacceptable performace. Therefore we have subsequently proposed, implemented, and tested new replication procedures and algorithms of unification and identities detection on real data sets.
These algortihms demonstrate huge improvements in many ways in operation conditions compared to the originally published ones.