System for recovering data stored in failed memory unit

• US 4,092,732 A
• Filed: 05/31/1977
• Issued: 05/30/1978
• Est. Priority Date: 05/31/1977
• Status: Expired due to Term

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Ouchi, Norman Ken
• Primary Examiner(s): Thomas, James D.
• Application Number:
  US05/801,961
• Time in Patent Office: 364 Days
• Field of Search: 364/200, 364/900, 235/301, 235/306, 235/312, 340/146.1 AJ
• US Class Current: 714/5.11
• CPC Class Codes: G06F 11/1076 Parity data used in redundant arrays of independent storages, e.g. in RAID systems

Abstract

A storage subsystem comprising a storage control unit and a number of failure independent storage devices, such as disk files, is provided with a check sum generator for generating a check sum segment from segments of a system record as the system record segments are being transferred between the storage subsystem and a CPU of a data processing system. The check sum segments and the associated system record segments are stored on different units so that if one unit containing a system record segment becomes unavailable, the unavailable segment is reconstructed during transfer of the other available segments and the check sum segment to the CPU.

Claims

1. In combination with a storage subsystem having a control unit and a plurality of disk storage disk storage drives connected to said control unit wherein said control unit includes control means for controlling the transfer of data in the form of segmented system records between a CPU of a data processing system and said storage subsystem through said control means in response to commands supplied to the control unit specifying CPU addresses of said system record, the length of said records, and the physical storage location in said disk drives where said segments are to be stored, the improvement comprising:

A. a generator for

(1) generating a check sum segment corresponding to the logical combination of all segments of a record simultaneously during transfer of a system record from said system to said drives, and

(2) recreating an unavailable record segment from other available records segments and said check sum segment simultaneously during a transfer of said available segments and said check sum segment from said drives to said system; and

B. circuitry under control of said control means including register means addressable by said control means for transferring

(a) signals to and from said genertor corresponding to segmented system records being transfered,

(b) a check sum signal to a selected drive; and

(c) a recreated unavailable system record segment to said CPU.

2. The combination recited in claim 1 in which said control means is a microprocessor.

3. The combination recited in claim 2 in which said generator includes a buffer, first means connecting the register means of said microprocessor to the input of said buffer, and second means selectively connecting the output of said buffer to said first means to permit the selective logical combination of successively transferred record segments.

4. The combination recited in claim 3 further including third means connecting the output of said buffer to said register means of said microprocessor to transfer the contents of said buffer to a drive or said CPU selectively under ocntrol of said microprocessor.

5. The combination recited in claim 4 in which said first means includes an Exclusive OR gate for logically combining the output of said buffer with record segment signals transferred to said generator from said CPU.

6. In a data processing system having a main memory for storing system records being processed, a plurality of failure independent storage drives for storing system records before and after processing and means for controlling the selective transfers of a predetermined system record between said main memory and said storage device, the improvement comprising:

an arrangement for storing said system records and an associated check sum to permit the recording of each system record stored on said storage devices when one of said devices becomes inoperable, said arrangement comprising;

means for dividing each system record into a plurality of equal segment said plurality being at least one less than the number of said storage devices;

means for transferring each of said segments of a system record between said main memory and a different one of said storage drives;

a check sum generator connected to said transfer means to generate a system record check sum for logically combining successively transferred segments of a record simultaneously as said segments are transferred to said devices; and

means for transferring said generated check sum segment from said generator to a still different device, whereby there is no more than one segment associated with a system record stored on one device.

7. The combination recited in claim 6 in which said arrangement includes control means responsive to control signals from said system for transferring said generated check sum to said still different device immediately following the transfer of the last segment of said system record.

8. The combination recited in claim 6 in which said arrangement further includes means responsive to the failure of one of said devices to transfer a segment of a requested system record, for recreating said unavailable segment with said check sum generator by transferring the associated store check sum to said genertor after the available segments have been transfered to said system.

9. The combination recited in claim 8 in which said arrangement includes control means responsive to said system for transferring said recreated segment to said main memory.

10. In a data processing system having a central processing unit with a main memory and a storage subsystem including a plurality of disk drives for storing system records to be processed and control means for transferring system records between assigned storage locations in said main memory and assigned storage locations in said disk drives, a method for generating a check sum for said system record simultaneously during transfer operations between said main memory and said drives which permits each system record stored in said subsystem to be reconstructed in said main memory when one of said disk drives becomes inoperable, said method comprising:

storing in main memory for each system record a plurality of addresses defining storage locations on different disk drives for storing equal length segments, at least one said address defining a location for storing a check sum segment and the other address defining locations for storing segments of said record predefined by said system;

transferring a system record to said storage devices serially by segment to store each said segment in a different disk drive determined by said stored other addresses;

12. The method recited in claim 10 further including the steps of:

transferring a system record serially by segments from said disk drives to said system and simultaneously generating during each of said segment transfers an intermediate check sum by logically combining each transferred segment with a temporality stored logically combined result of all prior transferred segment of that record;

recreating any single segment of said system record which is stored on an available drive by transferring said stored check sum segment from its drive and logically combining said check sum segment with said temporarily stored result; and

transferring the recreated segment to said system.

11. generating a system record check sum by logically combining during said above transfer step each transferred segment with a temporarily stored logically combined result of all prior transferred segments of that record;

and

transferring the final check sum from temporary storage to a disk drive address determined by said check sum address.