* News


Algorithms for Data Migration

Yoo-Ah Kim
University of Maryland, College Park

Monday, March 7, 2005
Ricketts 211 - 4:00 p.m. to 5:00 p.m.
Refreshments at 3:30 p.m.

Large storage systems typically consist of many devices storing data. To balance the load across disks, the system computes data layouts that are typically adjusted according to the workload. As workloads change over time, the system recomputes the data layout, and then rearranges the data items according to the new layout. We identify the problem of computing an efficient data migration plan that involves converting an initial layout to a target layout. We first show that this problem is NP-hard. We then develop approximation algorithms for the data migration problem and compare it to several other heuristic algorithms. In this talk we will present both the worst case algorithm, as well as the performance evaluation of the different methods through an experimental study. The talk will conclude with a brief overview of other research I have done, and a summary of my future research interests.

Last updated: February 24, 2005