Method and means for managing RAID 5 DASD arrays having RAID DASD arrays as logical devices thereof

• US 5,301,297 A
• Filed: 07/03/1991
• Issued: 04/05/1994
• Est. Priority Date: 07/03/1991
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Menon, Jaishankar M., Wood, Leighton C. Jr.
• Primary Examiner(s): Kulik, Paul V.
• Application Number:
  US07/725,696
• Time in Patent Office: 1,007 Days
• Field of Search: 395/425, 395/575, 371/7, 371/10.1, 371/10.2, 371/11.1, 371/40.1, 371/50.1
• US Class Current: 711/114
• CPC Class Codes: G06F 11/1076 Parity data used in redundant arrays of independent storages, e.g. in RAID systems
• CPC Class Codes: G06F 2211/1045 Nested RAID, i.e. implementing a RAID scheme in another RAID scheme

Abstract

A method and apparatus teaching insertion of addressing indirection to form and to access an array hierarchy expressly permitting the concurrency of a high level RAID array, the bandwidth and degraded mode operation sustainable by a lower level RAID array, and after a DASD failure minimum spanning involvement when the array is rebuilding and rewriting missing data to a spare logical device. Also, disclosed are the accessing of variable length records on the array hierarchy; array hierarchy in which RAID 5 arrays have dissimilar number of logic devices (lower level RAID arrays) and interleave depths; formation of logical arrays using fractional storage defined onto real DASD subsets; and the defining of logical devices onto DASDs distributed in the same or different physical clusters of DASDs and the rebuild operation thereof.

Claims

1. A method for accessing data strings from a plurality of DASDs, comprising the steps of:

(a) configuring the plurality of DASDs into a RAID 5 array of logical devices, each logical device including N data+P parity DASDs in RAID 3 DASD array;

(b) segmenting each data string into K logical records of N data blocks each, coding P parity blocks over each segmented record, and accessing said K logical records in column major order of predetermined interleave depth onto the RAID 5 array logical devices; and

(c) transferring K records at N data+P parity blocks at a time according to the order and depth from the RAID 3 DASD arrays.

2. A method for writing data strings applied to a storage subsystem, said subsystem being formed from a plurality of RAID 3 DASD arrays, each RAID 3 array having N data+P parity DASDs attaching a controller, comprising the steps of:

(a) configuring the plurality of RAID 3 DASD arrays into a RAID 5 array of addressable logical devices by communicatively attaching each of their controllers through a common control unit; and

(b) segmenting each data string at the common control unit into K logical records of N blocks each, parity coding each logical record and adding P blocks thereto, and

(c) writing K logical records of N data+P parity blocks, each in column major order of predetermined interleave depth, onto the RAID 5 array of logical devices such that the N data+P parity blocks of each logical record are recorded onto N data+P parity DASDs of a counterpart RAID 3 array operating as the addressed logical device through a path including its controller.

3. A method for managing large and small accesses of portions of data strings formed from K* (N data+P parity) blocks applied to a storage subsystem formed from a plurality of RAID 3 DASD arrays, each RAID 3 array having N data+P parity DASDs attaching a controller, comprising the steps of:

(a) configuring the plurality of RAID 3 DASD arrays into at least one level of indirection forming a RAID 5 array of logical devices of RAID 3 DASD arrays, and, reserving at least one of the logical devices as a spare device by communicatively attaching each controller to a common control unit;

(b) accessing the logical devices, other than the spare device, through a path including the common control unit in column major order of predetermined interleaved depth;

(c) detecting a DASD failure in a RAID 3 array;

(d) responsive to detection of said failure, accessing the logical devices as in step (b) but for the failed DASD; and

(e) rebuilding and rewriting N data+P parity DASDs of the logical device whose counterpart RAID 3 array contains the failed DASD onto a spare device on either a predetermined or opportunistic schedule, said rebuild and rewrite spanning only the spare device and said device having the failed DASD.

4. The method according to claim 3, wherein the accessing step further comprises:

(b1) accessing K* (N+P) blocks from the RAID 5 array in row major order K modulo M and in column major order K modulo M* (N+P), M lying in the closed integer interval (1,K), the partial sum N+P designating N data blocks+P parity blocks; and

(b2) executing large and small access requests over the RAID 5 array such that a minimum number X of blocks transferred to achieve the maximum data rate for a given M lies in the closed interval ((N+P-1, (N+P-1) *K) whose end points are defined by M=1 and M=K respectively.

5. The method according to claim 3, wherein the step of accessing all but the failed DASD in the RAID 3 array of an accessed logical device responsive to a small read access further includes:

reading all but the failed DASD and reconstructing any unavailable data block by logically combining the N+P-1 available blocks, the partial sum N+P designating N data blocks+P parity blocks; and

writing to all but the failed DASD such that the parity block and all but one of the data blocks are stored on the remaining DASDs of RAID 3 array of the accessed device.

6. A storage subsystem comprising:

a plurality of RAID 3 arrays of DASDs, each RAID 3 array including N data+P parity DASDs attaching a local controller;

a control unit communicatively attaching a subset of the plurality of RAID 3 arrays of DASDs to form an addressable RAID 5 array of logic device equivalents of the RAID 3 arrays through a path including the control unit and local controllers; and

means including the control unit and responsive to each data string formed from K*N blocks for segmenting the string into N blocks, encoding P parity blocks over each segment, and writing K of the N data+P parity block segments in column major order of predetermined interleave depth on to counterpart logical devices.

7. The storage subsystem according to claim 6, wherein said segmenting, encoding, and writing means further includes means for recording the N data+P parity blocks of each logical device onto N data+P parity DASDs of a counterpart RAID 3 array through a path including its controller.

8. A storage subsystem comprising:

a plurality of logical devices, each logic device including a RAID 3 array of DASDs having N data+P parity DASDs attaching a local controller;

a control unit communicatively attaching a subset of the plurality of logical devices to form an addressable RAID 5 array of said logical devices through a path including said control unit and local controllers, others of the plurality of logical devices being reserved as spare logical devices;

first means including the control unit responsive to each data string formed from K*N blocks for segmenting the string into N blocks, encoding P parity blocks over each segment, writing K of the N data+P parity block segments in column major order of a predetermined interleave depth on to counterpart logical devices, and for recording the N data+P parity blocks of each segment onto N data+P parity DASDs of a counterpart RAID 3 DASD array to the device through a path including its controller;

second means including the control unit for detecting a DASD failure in the RAID 3 DASD array counterpart of a logical device and for writing to the logical devices in a predetermined designated order and depth and recording the blocks but for those blocks destined for the failed DASD; and

third means for rebuilding and rewriting N data+P parity DASDs of the logical device whose counterpart RAID 3 array contains the failed DASD onto the spare logical device on either a predetermined or opportunistic schedule, said rebuild and rewrite spanning only the spare logical device and said logical device having the failed DASD.

9. The storage subsystem according to claim 8, wherein

said plurality of logical devices further comprising a first RAID 5 array of m logical devices and a second RAID 5 array of n logical devices, m not equal to n;

said control unit further comprises means for switchably coupling the control unit to any one of the RAID 5 arrays; and

said first means including the switching means further comprises means for selecting any of the RAID 5 arrays and recording segments on the logical devices thereof in column major order thereon to an interleave depth distinguishable from the interleave depth of at least one other RAID 5 array.

10. A storage subsystem comprising:

a plurality of logical devices, each device formed from a RAID 3 array of DASDs;

a control unit for communicatively attaching either a first or a second RAID 5 array of the logical devices;

first means including the control unit for selecting the first RAID 5 array, segmenting a data string into segments of m data blocks, coding p parity blocks over each segment, and writing each segment of m data+p parity blocks across the logical devices of the first RAID 5 array in column major order to a first interleave depth; and

second means including the control unit for selecting a second RAID 5 array, segmenting another data string into n data blocks, coding p parity blocks over each segment, and writing each segment of n data+p parity blocks across the second DASD subset in column major order to a second interleave depth, at least one of the devices of the second RAID 5 array being also a member of the first RAID 5 array.

11. The storage subsystem according to claim 10, wherein at least one logical device DASD of the plurality of logical devices reserving a fraction of its storage capacity to the first and the second RAID 5 arrays proportional to the number of logical devices in each RAID 5 array.

12. The storage subsystem according to claim 10, wherein n not being equal to m, and the second depth not being equal to the first depth.

13. A method for writing variable length, count-key-data (CKD) field formatted records to a storage subsystem, said subsystem being formed from a plurality of RAID 3 DASD arrays, each RAID 3 array having N data+P parity DASDs attaching a controller, comprising the steps of:

(a) configuring the plurality of RAID 3 DASD arrays into a RAID 5 array of addressable logical devices by communicatively attaching each of their controllers through a common control unit; and

(b) segmenting each data string at the common control unit into K logical records of N stripes each, parity coding each logical record and adding P strings thereto, and

(c) writing K logical records of N data+P parity strings, each in row major order of predetermined interleave depth, onto the RAID 5 array of logical devices such that the N data+P parity strings of each logical record are recorded onto N data+P parity DASDs of a counterpart RAID 3 array operating as an addressed logical device through a path including its controller such that;

(1) each count field of a CKD formatted variable length record being replicated on the N DASDs of the first device in row major order,

(2) the key and data fields following the count field of a CKD formatted variable length record being segmented and recorded across N DASDs of the counterpart RAID 3 array operating as the addressed logical device in the predetermined row major order and interleave depth, and

(3) each parity block P logically combining a counterpart N fields.