Check out the new USENIX Web site. next up previous
Next: Space Overheads Up: Improving Storage System Availability Previous: Fast Recovery


Exploring Graceful Degradation

In this section, we use simulation and trace analysis to evaluate the potential effectiveness of graceful degradation and the impact of different semantic grouping techniques. We first quantify the space overheads of D-GRAID. Then we demonstrate the ability of D-GRAID to provide continued access to a proportional fraction of meaningful data after arbitrary number of failures. More importantly, we then demonstrate how D-GRAID can hide failures from users by replicating ``important'' data. The simulations use file system traces collected from HP Labs [38], and cover 10 days of activity; there are 250 GB of data spread across 18 logical volumes.




Subsections

Muthian Sivathanu 2004-02-17