Permutation Development Data Layout (PDDL) Disk Array Declustering
Thomas Schwarz, Jesse Steinberg, Walter A. Burkhard · 1998
Declustered data organizations have been proposed to achieve less-intrusive reconstruction of a failed disk's contents. In previous work, Holland and Gibson identified six desirable properties for ideal layouts. Ideal layouts exist for a very limited family of configurations. The PRIME data layout deviates from the stated ideal only slightly and its run-time performance is very good for light workloads. The DATUM data layout does not meet one of the ideal layout properties but has very good run-time performance for heavy workloads. We present the Permutation Development Data Layout (PDDL) which has excellent run-time performance for both light and heavy workloads. Moreover, PDDL has straightforward space and run-time efficient implementations. Our simulation studies show that the new layouts provide excellent parallel access performance and reduced incremental loads during degraded operation. PDDL can easily accommodate multiple failure tolerant redundancy schemes.