Redistributing data across drives of a storage array based on drive health metrics
Techniques are presented for maintaining data distributed across a plurality of storage drives (drives) in a robust manner. A method includes (a) collecting physical state information from each drive of the plurality of drives, (b) generating a predicted failure probability of each drive based on the collected physical state information from that drive, the predicted failure probability indicating a likelihood that that drive will fail within a predetermined period of time, and (c) rearranging a distribution of data across the plurality of drives to minimize a probability of DU/DL. Systems, apparatuses, and computer program products for performing similar methods are also provided.
1. A method of maintaining data distributed across a plurality of storage drives (drives) in a robust manner, the method comprising:
collecting physical state information from each drive of the plurality of drives;
generating a predicted failure probability of each drive based on the collected physical state information from that drive, the predicted failure probability indicating a likelihood that that drive will fail within a predetermined period of time; and
rearranging a distribution of data across the plurality of drives to minimize a probability of data unavailability/data loss (DU/DL) by remapping an address space of a logical disk (LUN) so that areas of drives that have lower predicted failure probabilities are configured to store user data and areas of drives that have higher predicted failure probabilities are configured to map to areas of the address space that do not store user data.
2. The method of claim 1 , wherein the plurality of drives are arranged in a fixed RAID configuration having a set of stripes, each stripe going across all drives of the plurality of drives; and remapping the address space includes mapping the address space of the LUN to exclude areas of stripes that are on drives that have higher predicted failure probabilities.
3. The method of claim 2 wherein mapping the address space of the LUN to exclude areas of stripes that are on drives that have higher failure probabilities includes moving data already stored on the areas of stripes that are on drives that have higher failure probabilities to areas of stripes that are on drives that have lower failure probabilities.
4. The method of claim 2 wherein mapping the address space of the LUN to exclude areas of stripes that are on drives that have higher failure probabilities includes writing new user data to areas of stripes that are on drives that have lower failure probabilities instead of to areas of stripes that are on drives that have higher failure probabilities.
5. The method of claim 1 , wherein the plurality of drives are arranged in a Mapped RAID configuration having a set of stripes, each stripe going across fewer than all of the plurality of drives; and remapping the address space includes mapping the address space of the LUN to exclude areas of stripes that are on drives that have higher predicted failure probabilities.
6. The method of claim 1 ,
wherein the plurality of drives are arranged in a Mapped RAID configuration having a set of stripes, each stripe going across fewer than all of the plurality of drives; and
wherein rearranging the distribution of data includes rearranging the set of stripes so that areas of stripes that are initially on drives that have higher predicted failure probabilities are moved to drives that have lower predicted failure probabilities.
7. The method of claim 1 wherein rearranging the distribution of data includes:
determining a current probability of DU/DL;
searching a space of different distributions of data across the plurality of drives for a distribution of data that has a lowest probability of DU/DL; and
in response to determining that the current probability of DU/DL exceeds the lowest probability of DU/DL by at least a predetermined threshold, moving data in accordance with the distribution of data that has the lowest probability of DU/DL.
8. The method of claim 1 wherein rearranging the distribution of data includes:
determining a current probability of DU/DL;
searching a space of different distributions of data across the plurality of drives for an alternate distribution of data that satisfies the following conditions:
the alternate distribution has a lower probability of DU/DL such that the current probability of DU/DL exceeds the lower probability of DU/DL by at least a predetermined threshold and
the alternate distribution has a skew below a maximum permitted skew value; and
moving data in accordance with the alternate distribution of data.
9. A computer program product comprising a non-transitory computer-readable storage medium storing a set of instructions, which, when performed by a computing device, cause the computing device to maintain data distributed across a plurality of storage drives (drives) in a robust manner by:
collecting physical state information from each drive of the plurality of drives; generating a predicted failure probability of each drive based on the collected physical state information from that drive, the predicted failure probability indicating a likelihood that that drive will fail within a predetermined period of time; and
rearranging a distribution of data across the plurality of drives to minimize a probability of data unavailability/data loss (DU/DL) by remapping an address space of a logical disk (LUN) so that areas of drives that have lower predicted failure probabilities are configured to store user data and areas of drives that have higher predicted failure probabilities are configured to map to areas of the address space that do not store user data.
10. The computer program product of claim 9 ,
wherein the plurality of drives are arranged in a Mapped RAID configuration having a set of stripes, each stripe going across fewer than all of the plurality of drives; and
wherein rearranging the distribution of data includes rearranging the set of stripes so that areas of stripes that are initially on drives that have higher predicted failure probabilities are moved to drives that have lower predicted failure probabilities.
11. A data storage system (DSS) comprising:
a plurality of storage drives (drives);
a set of persistent storage device that provide a pool of persistent storage; and
a computing device communicatively-connected to the set of persistent storage devices, the computing device being configured to maintain data distributed across the plurality of drives in a robust manner by:
collecting physical state information from each drive of the plurality of drives;
generating a predicted failure probability of each drive based on the collected physical state information from that drive, the predicted failure probability indicating a likelihood that that drive will fail within a predetermined period of time; and
rearranging a distribution of data across the plurality of drives to minimize a probability of data unavailability/data loss (DU/DL) by remapping an address space of a logical disk (LUN) so that areas of drives that have lower predicted failure probabilities are configured to store user data and areas of drives that have higher predicted failure probabilities are configured to map to areas of the address space that do not store user data.
12. The DSS of claim 11 ,
wherein the plurality of drives are arranged in a Mapped RAID configuration having a set of stripes, each stripe going across fewer than all of the plurality of drives; and
wherein rearranging the distribution of data includes rearranging the set of stripes so that areas of stripes that are initially on drives that have higher predicted failure probabilities are moved to drives that have lower predicted failure probabilities.