System and method for improved rebuild in RAID

• US 8,751,861 B2
• Filed: 02/08/2012
• Issued: 06/10/2014
• Est. Priority Date: 02/08/2012
• Status: Active Grant

• Current Assignee: Avago Technologies International Sales Pte Limited (Broadcom, Inc.)
• Original Assignee: LSI Corporation (Broadcom, Inc.)
• Inventors: Nair, Naman, Le, Khai M.
• Primary Examiner(s): DUNCAN, MARC M
• Application Number:
  US13/368,725
• Publication Number:
  US 20130205166A1
• Time in Patent Office: 853 Days
• Field of Search: 714/6.2, 714/6.21, 714/6.22
• US Class Current: 714/6.21
• CPC Class Codes: G06F 11/1092 Rebuilding, e.g. when physically replacing a failing disk
• CPC Class Codes: G06F 3/0611 in relation to response time
• CPC Class Codes: G06F 3/0613 in relation to throughput
• CPC Class Codes: G06F 3/0659 Command handling arrangements, e.g. command buffers, queues, command scheduling
• CPC Class Codes: G06F 3/0689 Disk arrays, e.g. RAID, JBOD

Abstract

The present disclosure is a system and method for improved RAID rebuilds under host IO conditions, that greatly improves rebuild times and prevents host IO starvation. A queue in a drive that is part of the RAID is used to store rebuild and host IO requests, with rebuild IOs issued to the head of the drive queue. Rebuild requests in the drive are delayed by a delay time. This delay ensures there is no unintended side effect of this invention that may result in host IO starvation for the RAID system. Delay is computed as a heuristic function from a plurality of variables, such as disk head response times, time allotted for rebuild, number of rebuild threads, drive behavior, rebuild IO profile, host IO profile including workload and locality, and a timeline of sequenced events, which in one preferred embodiment cycles every 70 seconds.

Claims

1. A storage system having a plurality of drives for improved RAID rebuilds, comprising:

firmware to determine an average rebuild IO response time of a drive in the storage system;

a queue internal to the drive comprising rebuild IOs and host IOs requests;

firmware to delay issuance of rebuild IOs to the queue;

firmware to calibrate the delay according to the parameters of average rebuild response times from the queue, a time allotted for rebuild compared to a total time and a number of rebuild threads; and

firmware to issue rebuild IOs to the queue using a SCSI IO tag head-of-queue if the average rebuild response time exceeds a threshold.

2. The invention according to claim 1, further comprising:

firmware having a rebuild rate that specifies how quickly the firmware is to rebuild, the rate having a minimum and a maximum value, with host IOs prioritized more over rebuild IOs at the minimum value versus the maximum value.

3. The invention according to claim 2, wherein:

the rebuild rate in the firmware being configurable, allowing the firmware to have rebuild rates that vary between several predetermined values ranging from the minimum to the maximum values.

4. The invention according to claim 1 wherein:

the firmware to calibrate the delay is configured to calibrate the delay according to further parameters selected from the group consisting of read response times, time allotted for rebuild, user-selected Rebuild IO Profile, and configurable Rebuild IO Profile.

5. The invention according to claim 4, further comprising:

firmware to issue rebuild IOs to the head of the queue in a periodic manner, having a plurality of stages comprising the period, comprising a first stage that proceeds to rebuild by issuing rebuild IOs in a traditional manner in the queue, without the introduction of the added delay, a second stage that is reached when at least one of a rebuild rate and response time are less than predetermined values for rebuild rate and response time, with the computation of the delay T occurring during the second stage, and a third stage for issuing rebuild IOs to the queue at the front of the queue every T ms.

6. The invention according to claim 5, wherein:

the second stage computes the delay T according to the equation;

T=C/nR−

resp

where;

nR=rIOPS/num rebuild threads

rIOPS=rt/resp

where;

“

num rebuild threads”

=number of rebuild threads in the RAID system,

rt=the time allotted for rebuild compared to the total time (C),

resp=the average head of queue rebuild read response times at the head of the queue, and,

C=a time constant, the total time.

7. The invention according to claim 5, wherein:

the firmware operates in a storage system selected from the group consisting of RAID 1, RAID 2, RAID 3, RAID 4, RAID 5, RAID 6, hybrid RAID levels 1 through 6 that include RAID 0, storage systems having copyback operations, and storage using erasure codes.

8. A method for RAID rebuilds for a plurality of drives in a storage system comprising the steps of:

determining average rebuild IO response times in a RAID system operating rebuilds;

computing a delay by which to delay issuance of a rebuild IO to the head of queue of a drive in the RAID according to parameters of average rebuild response times from the queue, a time allotted for rebuild compared to total time and the number of rebuild threads; and

,

issuing rebuild IOs to the queue using a SCSI IO tag head-of-queue if the average rebuild response time exceeds a threshold and delaying execution of the rebuild IOs by the delay.

9. The method according to claim 8,

wherein the delay is computed according to further parameters selected from the group consisting of read response times, time allotted for rebuild, user-selected Rebuild IO Profile, and configurable Rebuild IO Profile.

10. The method according to claim 8, further comprising the steps of:

selecting the time allotted for rebuild from a minimal rebuild rate to a maximum rebuild rate.

11. The method according to claim 10, further comprising the steps of:

selecting the time allotted for rebuild from a configurable Rebuild Profile, the selection of time allotted for rebuild in the Rebuild Profile comprising a plurality of rebuild rates ranging from the minimal rebuild rate to the maximum rebuild rate.

12. The method according to claim 8,

wherein the delay is computed according to further parameters comprising read response times, time allotted for rebuild, user-selected Rebuild IO Profile, and configurable Rebuild IO Profile.

13. The method according to claim 12, wherein:

the steps of determining average rebuild IO response times, computing delay for the queue, and issuing rebuild IOs to the queue occurs in a periodic manner in the RAID storage system.

14. The method according to claim 13, wherein:

the periodic manner of steps comprises a first stage which proceeds to rebuild the RAID storage system by issuing rebuild IOs in a traditional manner in the queue, without issuing the rebuild IOs to the head of drive queue, a second stage that is reached when at least one of a rebuild rate and response time are greater than predetermined values for rebuild rate and response time, with the computation of the delay T in ms occurring during the second stage, and a third stage for issuing rebuild IOs to the head of drive queue every T ms.

15. The method according to claim 14, wherein:

the method is used in a storage system selected from the group consisting of RAID 1, RAID 2, RAID 3, RAID 4, RAID 5, RAID 6, hybrid RAID levels 1 through 6 that include RAID 0, storage systems having copyback operations, and storage using erasure codes, storage systems employing issuing IOs to physical drives, storage systems involving media scans, storage systems involving consistency checks, storage systems involving initializations, and storage systems involving formatting.

16. An apparatus for RAID rebuilds for RAID rebuilds under host IO conditions comprising:

means for measuring rebuild IO responses in a RAID storage system;

means for queuing for storing rebuild IO and host IO requests for said RAID storage system;

means for computing a delay to delay issuance of said rebuild IO requests to the head of said queue comprising a delay time in milliseconds;

means for issuing rebuild IOs to head of said queue and delaying the execution of the rebuild IOs by said delay time;

means for setting a rebuild rate for said RAID system;

said RAID storage system means for measuring rebuild IO responses, means for computing delay and means for issuing rebuild IOs operate in a periodic manner, wherein,

said means for measuring rebuild IO responses in a first time period computes the average rebuild response times of the RAID system when operating in a traditional manner, until such time either one of the response rate of the RAID system and the response times of the RAID system rise above predetermined values;

said means for computing a delay in a second time period computes the optimal delay by computing the average rebuild response time when said RAID system is operating with said delay instructions added to head of said drive queue, the rebuild rate of said means for setting a rebuild rate, and the number of threads in said RAID system; and

,

said means for issuing rebuild IOs delays issuance of a rebuild IO request in a third time period to head of said drive queue by said optimal delay computed in said second time.

17. The apparatus according to claim 16, wherein:

said means for computing a delay in a second time period computes the optimal delay T according to the equation;

T=C/nR−

resp

where;

nR=rIOPS/num rebuild threads

rIOPS=rt/resp

where;

“

num rebuild threads”

=number of rebuild threads in the RAID system,

rt=the time allotted for rebuild compared to the total time (C),

resp=the average head of queue rebuild read response times at the head of queue, and,

C=a time constant, the total time.