Automated identification of the source of RAID performance degradation

Patent metadata

Field	Value
Publication number	US-9892014-B1
Application number	US-201414500485-A
Country	US
Kind code	B1
Filing date	Sep 29, 2014
Priority date	Sep 29, 2014
Publication date	Feb 13, 2018
Grant date	Feb 13, 2018

Abstract

Exemplary methods for determining performance degradation of one or more disks at a storage system include analyzing disk performance statistics of a target storage system periodically collected during a storage of data from a source storage system to the target storage system, wherein the target storage system is a redundant array of inexpensive disks (RAID) storage system comprising of one or more arrays, each array comprising of a plurality of disks. The methods further include creating a predictive model based on the analysis of the collected performance statistics, and predicting, using the predictive model, that one or more disks of an array of the one or more arrays of the RAID storage system has degraded in performance.

First claim

What is claimed is:

1. A computer-implemented method, the method comprising:
analyzing disk performance statistics of a target storage system periodically collected during a storage of data from a source storage system to the target storage system, wherein the target storage system is a redundant array of inexpensive disks (RAID) storage system comprising of one or more arrays, each array comprising of a plurality of disks, wherein the disk performance statistics include at least a usage observation;

creating a predictive model based on the analysis of the collected performance statistics, wherein creating the predictive model includes:
determining a population mean of the disk performance statistics, and
determining a population standard deviation of the disk performance statistics based on the population mean; and
predicting, using the predictive model, that one or more disks of an array of the one or more arrays of the RAID storage system has degraded in performance, wherein predicting, using the predictive model, that the one or more disks of the array has degraded includes:
generating a Z-Score for each of the one or more disks of the array, wherein, for each of the one or more disks, generating the Z-score includes:
determining a difference between the usage observation and the population mean of usage observations, and dividing the difference by the population standard deviation of the usage observations, and

determining, for each of the one or more disks, whether the Z-Score is greater than a Z-Score threshold, wherein, when the Z-Score is greater than the Z-Score threshold, a disk degradation is predicted.

2. The method of claim 1 , wherein the disk performance statistics include the usage observation for each disk of the array of the RAID storage system, wherein each usage observation indicates an amount of time a respective disk was busy processing disk access requests during a time interval.

3. The method of claim 2 , wherein creating the predictive model comprises:
determining the population mean of the usage observations associated with the plurality of disks of the array the RAID storage system; and
determining the population standard deviation of the usage observations associated with the plurality of disks of the array of the RAID storage system based on the population mean.

4. The method of claim 3 , wherein predicting that one or more disks of the array of the RAID storage system has degraded in performance comprises:
determining that the usage observation of one or more disks of the array of the RAID storage system is at least a predetermined number of standard deviations from the population standard deviation.

5. The method of claim 4 , wherein the disk performance statistics further include a read byte count and a write byte count for each disk of the array of the RAID storage system, wherein each read byte count indicates a number of bytes read from a respective disk during the time interval, and wherein each write byte count indicates a number of bytes written to a respective disk during the time interval.

6. The method of claim 5 , wherein predicting that one or more disks of the array of the RAID storage system has degraded in performance further comprises:
determining the population mean of the read byte counts associated with the plurality of disks of the array of the RAID storage system;
determining the population standard deviation of the read byte counts associated with the plurality of disks of the array of the RAID storage system based on the population mean;
determining the population mean of the write byte counts associated with the plurality of disks of the array of the RAID storage system; and
determining the population standard deviation of the write byte counts associated with the plurality of disks of the array of the RAID storage system based on the population mean.

7. The method of claim 6 , wherein predicting that one or more disks of the RAID storage system has degraded in performance further comprises:
determining a standard deviation sum of the population standard deviation of the read byte counts and the population standard deviation of the write byte counts; and
determining that the standard deviation sum is within a predetermined percentage of the population standard deviation of the usage observations.

8. The method of claim 4 , wherein predicting that one or more disks of the array of the RAID storage system has degraded in performance further comprises:
determining that the usage observation of one or more disks of the array of the RAID storage system is at least the predetermined number of standard deviations from the population standard deviation for at least a first predetermined amount of time out of a second consecutive predetermined amount of time.

9. The method of claim 4 , wherein predicting that one or more disks of the array of the RAID storage system has degraded in performance further comprises:
determining that the usage observation of one or more disks of the array of the RAID storage system is greater than a predetermined usage threshold.

10. A non-transitory computer-readable storage medium having computer code stored therein, which when executed by a processor, cause the processor to perform operations comprising:
analyzing disk performance statistics of a target storage system periodically collected during a storage of data from a source storage system to the target storage system, wherein the target storage system is a redundant array of inexpensive disks (RAID) storage system comprising of one or more arrays, each array comprising of a plurality of disks, wherein the disk performance statistics include at least a usage observation;

creating a predictive model based on the analysis of the collected performance statistics, wherein creating the predictive model includes:
determining a population mean of the disk performance statistics, and
determining a population standard deviation of the disk performance statistics based on the population mean; and
predicting, using the predictive model, that one or more disks of an array of the one or more arrays of the RAID storage system has degraded in performance, wherein predicting, using the predictive model, that the one or more disks of the array has degraded includes:
generating a Z-Score for each of the one or more disks of the array, wherein, for each of the one or more disks, generating the Z-score includes:
determining a difference between the usage observation and the population mean of usage observations, and dividing the difference by the population standard deviation of the usage observations, and

determining, for each of the one or more disks, whether the Z-Score is greater than a Z-Score threshold, wherein, when the Z-Score is greater than the Z-Score threshold, a disk degradation is predicted.

11. The non-transitory computer-readable storage medium of claim 10 , wherein the disk performance statistics include the usage observation for each disk of the array of the RAID storage system, wherein each usage observation indicates an amount of time a respective disk was busy processing disk access requests during a time interval.

12. The non-transitory computer-readable storage medium of claim 11 , wherein creating the predictive model comprises:
determining the population mean of the usage observations associated with the plurality of disks of the array the RAID storage system; and
determining the population standard deviation of the usage observations associated with the plurality of disks of the array of the RAID storage system based on the population mean.

13. The non-transitory comput

Assignees

• Emc Corp
• Emc Ip Holding Co Llc

Inventors

• Hickey Charles
• Chamness Mark
• Levin Michael Orit

Classifications

• G06F3/0689

  Disk arrays, e.g. RAID, JBOD · CPC title

• G06F11/3452 Primary

  Performance evaluation by statistical analysis · CPC title

• G06F11/008 Primary

  Reliability or availability analysis · CPC title

• G06F3/0616

  in relation to life time, e.g. increasing Mean Time Between Failures [MTBF] · CPC title

• G06F3/0647

  Migration mechanisms · CPC title