Tesis doctoral de Francisco Javier Hidrobo Torres
In this thesis, we presented a system able to improve the i/o performance in an automatic way. In order to obtain the autonomic operation, we propose a system able to predict the behavior of the applications with regard to files access pattern. On the other hand, the system is able to propose dierent data placements and to predict the performance that will be obtained with these new distributions. Finally, with this information the system is able to decide if exist some better distribution and therefore to implement it. this system is build by three independent modules. The first one is able to learn the behavior of a workload in order to reproduce its behavior later on, without a new execution, even when the workload behavior change slightly or data placement is modified. The second module is a drive modeler that is able to learn how a storage drive works in an automatic way, just executing some synthetic tests once the disk is initially installed. Finally, the third module generates a set of placement alternatives and uses previous modules to predict the performance each alternative will achieve. we tested the system with synthetic and real workloads, in all cases the system was able to detect when were necessary or not to make some change to improve the i/o performance. And, more important, the system predictions were precise and the dierence between real behavior and predicted behavior was below 10%.
Datos académicos de la tesis doctoral «Sistema de almacenamiento autonomico basado en aprendizaje automatico«
- Título de la tesis: Sistema de almacenamiento autonomico basado en aprendizaje automatico
- Autor: Francisco Javier Hidrobo Torres
- Universidad: Politécnica de catalunya
- Fecha de lectura de la tesis: 28/05/2004
Dirección y tribunal
- Director de la tesis
- Antonio Cortés Roselló
- Tribunal
- Presidente del tribunal: Jesús josé Labarta mancho
- vicente Santonja (vocal)
- Jesús Carretero pérez (vocal)
- Manuel Garcia jose (vocal)