System for rebuilding dispersed data

  • US 7,546,427 B2
  • Filed: 04/13/2006
  • Issued: 06/09/2009
  • Est. Priority Date: 09/30/2005
  • Status: Active Grant
First Claim
Patent Images

1. A method comprising the steps of:

  • (a) assembling a list of unavailable storage nodes;

    (b) compiling a list of affected data segments contained on the unavailable storage nodes;

    (c) determining a list of data slices needed to rebuild the affected data segments;

    (d) reading the data slices needed to rebuild the affected data segments from associated storage nodes;

    (e) rebuilding the affected data segments from the data slices;

    (f) creating new data slices from the rebuilt data segment using an information dispersal algorithm; and

    (g) writing the new data slices associated with the unavailable storage nodes to different storage nodes;

    wherein the step of creating new data slices comprises partitioning a data segment into a set of data slices, and creating an encoded data subset by summing a modulo of at least two data slices within a collection of data slices; and

    wherein the step of writing further includes storing each data slice along with its corresponding encoded data subset in the storage node; and

    wherein the step of rebuilding the data segments comprises combining a subset of the data slices and the encoded data subsets required to reproduce the data segment from less than all of the original data slices.

View all claims

    Thank you for your feedback