P.T. Breuer (Spain) and A. Wiebalck (Germany)
Fault-tolerant Mass Storage, Networked RAID, Distributed RAID, Fast Resynchronization.
This article sets out an architecture for RAID over the net work, in the context of storage for large computational grids. In the target setting, "network block devices" substi tute for the hard disk drives found in a standard RAID array. The architecture incorporates special mechanisms for re porting, recording and recovery which lower transfer totals during RAID re-synchronisation, lower latency, and pro vide an intelligent control that recognises and compensates automatically for temporary network brown-outs. The ar chitecture has been implemented as a set of kernel drivers for the Linux operating system, and preliminary benchmark results for the prototype are presented here.
Important Links:
Go Back
