J. Perez O., R.A. Pazos R., G. RodrÃguez, J. Frausto S., L. Cruz R., and H. Fraire H. (Mexico)
Data Allocation, Data Replication, Data Mobility.
This paper presents an extension of the DFAR mathematical optimization model, which unifies the allocation and dynamic migration of data-objects in distributed systems. The extension consists of the modification of the objective function and the addition of two constraints that model the replication of data-objects in network sites. This approach differs from the traditional methods that treat allocation and replication in two separate phases, in that the model solves both simultaneously. The model uses current state information of the system and usage statistical data collected during a given period, and adapts the system schema to the new users' location and usage patterns so as to minimize communication costs. Implicitly the model handles mobility and temporality. The replication model is particularly important for increasing the availability of data-objects in mobile computation and for improving the access to large data-objects in distributed systems. The Threshold Accepting Algorithm is a variation of the heuristic method known as Simulated Annealing, and it was used for solving the model. The paper includes experimental results obtained for large test cases.
Important Links:
Go Back
