Highly available fault tolerant relocation of storage with atomicity

• US 5,504,857 A
• Filed: 03/04/1993
• Issued: 04/02/1996
• Est. Priority Date: 06/08/1990
• Status: Expired due to Fees

• Current Assignee: International Business Machines SA (International Business Machines Corporation)
• Original Assignee: International Business Machines SA (International Business Machines Corporation)
• Inventors: Myers, James J., Eisenberger, George, Tetzlaff, William H., Bozman, Gerald P., Lett, Alexander S., Kamerman, Albert, McAssey, John J., Baird, Robert, Wang, Pong-sheng
• Primary Examiner(s): Beausoliel, Jr., Robert W.
• Assistant Examiner(s): Snyder, Glenn
• Application Number:
  US08/026,278
• Time in Patent Office: 1,125 Days
• Field of Search: 395/575, 395/182.04, 364/200, 364/222.82, 364/266.5, 364/268.5, 364/268.9, 364/285.3, 371/10.1
• US Class Current: 714/6.1
• CPC Class Codes: G06F 11/1402 Saving, restoring, recovering or retrying
• CPC Class Codes: G06F 12/08 in hierarchically structured memory systems, e.g. virtual memory systems

Abstract

A computer implemented method for robustly copying pages to system managed storage in order to maintain data in a consistent state and in order to provide continuous access availability of the pages to executing applications. The method achieves data consistency by atomically shadow copying application referenced pages and amending directories in a failure independent medium on (1) an access path interrupt as well as on (2) a page update basis. Availability is enhanced by duplexing the pages and directories as part of the atomic shadow copying step.

Claims

1. A CPU implemented method for maintaining consistency among pages or groups of related pages resident on failure independent devices of system managed storage (SMS) and for enhancing page or page group availability to CPU executing applications, said CPU having directory lockable access paths to the pages resident on SMS devices subject to interference or fault, executing applications causing page read and write accesses to be made through the operating system to SMS, said SMS including randomly accessed internal storage and directly accessed external storage, pages or groups of pages being accessible by way of a first logical address space mappable into a second logical address space, and lastly mappable from the second logical address space into an absolute location on a failure independent device, wherein the improvement comprises:

(a) responsive to each write access to a first logical address creating a page or group of pages, the step of generating and recording a pair of address tuples indicative of the location of said created pages or group of pages in said first and second logical address and absolute address spaces in a first directory, each address tuple correlating a second logical with an absolute address and a first logical with a second logical address for each page or group of pages, said first directory being referenced to establish access paths to the pages, and writing said page or group at the absolute address of a device in SMS; and

(b) responsive to a write update access to the first logical space address of an SMS stored page or group of pages, or, responsive to an interrupt in the access path, the step of shadow writing the page or group onto another device and atomically generating and recording another pair of address tuples (address in the second logical space, absolute address;

address in the first logical space, address in the second logical space) in said first directory.

2. The method according to claim 1, wherein the step responsive to a write update access or to an interrupt in the access path further includes the step of executing a step selected from the set consisting of (a) writing of a single page in place, (b) writing of a group of related pages in place by their address in the second logical space, and (c) relocation of pages within external storage.

3. A method for robustly copying of pages within a CPU storage subsystem having failure independent storage media substituents thereof, said copying ensuring subsystem availability of pages accessed by an executing application, wherein the improvement comprises:

(a) responsive to each write operation creating a page or group of related pages addressable in a first logical address space, the steps of defining a first concordance of a address for the page or group in a second logical address space, and an address in the storage subsystem for said page or group in an absolute address space, defining a second concordance of said first logical address and counterpart second logical address for the page or group, and writing the page at the absolute address on a first failure independent device in the storage subsystem; and

(b) responsive either to each write update or to each interrupt in establishing an access path to a page or group in the storage subsystem addressable by a counterpart address in the first logical address space, the steps of copying the page or group having said counterpart address in said first logical address space from its current addresses in the second logical address and absolute address spaces to another failure independent device in the storage subsystem at new addresses in the second logical and absolute address spaces, and, atomically updating the concordances such that the new address of the page or group in the second logical address space is paired with the new address in the absolute address space, and the address of the page or group in the first logical address space is paired with the new address in the second logical address space.

4. In a system having a processor having an internal memory, an operating system, a data cache formed from processor internal memory, an external store coupling said data cache and including failure independent storage devices, and means responsive to application generated read and write ops issued via the operating system for accessing a page or group of related pages either resident in the cache or as refreshed into cache from the external store, said means for accessing a page or group of related pages utilizing directory lockable paths to a page or group, the path for each page or group including an address in a first logical address space mapped to an address in a second logical address space, and where the address in the second logical address space is mapped to an absolute storage subsystem location (device, offset) on a first storage device,

said processor having means including the operating system for providing interrupt indication of access path fault, interference or degraded performance;

wherein the improvement comprises;

means for dynamically defining and updating page or group location directories for the external store as absolute addresses indexed by addresses in the second logical address space and as addresses in the second logical address space indexed by addresses in the first logical address space;

means responsive to each interrupt or page update or group update accessed by an address in the first logical address space for shadow copying the addressed page or group on a second storage device at a new address in the second logical address space and a new absolute address; and

means for atomically changing the directories including the entries in which the new absolute address is indexed by the new address in the second logical address space and the new address in the second logical address space is indexed by a counterpart address in the first logical address space.