US 11,809,727 B1
Predicting failures in a storage system that includes a plurality of storage devices
Frank Tuzzolino, San Jose, CA (US); John Colgrove, Los Altos, CA (US); Taher Vohra, Sunnyvale, CA (US); Andrew Kleinerman, Belmont, CA (US); Xiaohui Wang, San Jose, CA (US); and Benjamin Scholbrock, Sunnyvale, CA (US)
Assigned to PURE STORAGE, INC., Santa Clara, CA (US)
Filed by Pure Storage, Inc., Mountain View, CA (US)
Filed on Apr. 29, 2018, as Appl. No. 15/965,961.
Application 15/965,961 is a continuation in part of application No. 15/786,187, filed on Oct. 17, 2017, granted, now 10,564,884.
Application 15/786,187 is a continuation of application No. 15/140,445, filed on Apr. 27, 2016, granted, now 9,841,921, issued on Dec. 12, 2017.
Int. Cl. G06F 3/06 (2006.01); G11C 11/4074 (2006.01); G11C 11/4096 (2006.01)
CPC G06F 3/0647 (2013.01) [G06F 3/067 (2013.01); G06F 3/0607 (2013.01); G06F 3/0683 (2013.01); G11C 11/4074 (2013.01); G11C 11/4096 (2013.01)] 15 Claims
OG exemplary drawing
 
1. A method comprising:
gathering, from headers attached to pages within each of a plurality of blocks within a plurality of storage devices in a storage system, information describing a previous state of the plurality of blocks;
training a machine learning block lifespan model using the information, wherein the information identifies characteristics of one or more previously failed blocks of the plurality of blocks as well as characteristics of one or more other blocks of the plurality of blocks that have not previously failed;
determining, based on the information describing the previous state of the plurality of blocks within the storage devices and the machine learning block lifespan model, a predicted lifespan for at least one block, of the plurality of blocks, that stores data;
detecting that the predicted lifespan for the at least one block is below a predetermined block lifespan threshold; and
based on the detection, migrating the data stored in the at least one block therefrom.