REGISTER  |  SIGN IN
United States Patent6044444
OfekMarch 28, 2000
Title
Remote data mirroring having preselection of automatic recovery or intervention required when a disruption is detected
Abstract
Two data storage systems are interconnected by a data link for remote mirroring of data. Each volume of data is configured as local, primary in a remotely mirrored volume pair, or secondary in a remotely mirrored volume pair. Normally, a host computer directly accesses either a local or a primary volume, and data written to a primary volume is automatically sent over the link to a corresponding secondary volume. Each remotely mirrored volume pair can operate in a selected synchronization mode including synchronous, semi-synchronous, adaptive copy--remote write pending, and adaptive copy--disk. A number of automatic and non-automatic recovery mechanisms are provided on a logical volume basis for a desired level of data integrity and degree of operator or application program involvement. If a "volume domino" mode is enabled for a remotely mirrored volume pair, access to a volume of the pair is denied when the other volume is inaccessible. In a "links domino" mode, access to all remotely mirrored volumes is denied when remote mirroring is disrupted by an all-links failure. The domino modes can be used to initiate application-based recovery, for example, recovering a secondary data file using a secondary log file. In an overwrite cache mode, remote write-pending data in cache can be overwritten. In a "log-in-cache" mode, multiple versions of write data can be stored in cache in order to recover from a "rolling disaster".
Inventors:Ofek; Yuval (Hopkinton, MA)
Assignee:EMC Corporation (Hopkinton, MA)
Appl. No.:819672
Filed:March 17, 1997
Current U.S. Class:711/162 714/5 714/6 714/718 714/763 710/1 711/161 
Current International Class:G06F 11/20 (20060101)
Field of Search:711/162,144,145,161 707/6,7 395/182 371/40 710/1 714/5,6,718,763
U.S. Patent Documents
3588839June 1971Belady et al.
3835260September 1974Prescher et al.
3866182February 1975Yamada et al.
4020466April 1977Cordi et al.
4057849November 1977Ying et al.
4084231April 1978Capozzi et al.
4094000June 1978Brudevold
4124843November 1978Bramson et al.
4150429April 1979Ying
4161777July 1979Behnase et al.
4204251May 1980Brudevold
4342079July 1982Stewart et al.
4396984August 1983Videki, II
4430727February 1984Moore et al.
4453215June 1984Reid
4464713August 1984Benhase et al.
4577272March 1986Ballew et al.
4608688August 1986Hanson et al.
4634100January 1987Julich et al.
4698808October 1987Ishii
4710870December 1987Blackwell et al.
4755928July 1988Johnson et al.
4769764September 1988Levanon
4779189October 1988Legvold et al.
4783834November 1988Anderson et al.
4785472November 1988Shapiro
4797750January 1989Karweit
4805106February 1989Pfeifer
4814592March 1989Bradt et al.
4837680June 1989Crockett et al.
4841475June 1989Ishizuka
4849978July 1989Dishon et al.
4862411August 1989Dishon et al.
4888686December 1989Sinz et al.
4916605April 1990Beardsley et al.
4949187August 1990Cohen
4985695January 1991Wilkinson et al.
5007053April 1991Iyer et al.
5029199July 1991Jones et al.
5051887September 1991Berger et al.
5060142October 1991Menon et al.
5060185October 1991Naito et al.
5089958February 1992Horton et al.
5097439March 1992Partriquin et al.
5099485March 1992Bruckert et al.
5123099June 1992Shibata et al.
5127048June 1992Press et al.
5132787July 1992Omi et al.
5134711July 1992Asthana et al.
5146576September 1992Beardsley et al.
5146605September 1992Beukema et al.
5155814October 1992Beardsley et al.
5155835October 1992Belsan
5155845October 1992Beal et al.
5157770October 1992Beardsley et al.
5159671October 1992Iwami
5170471December 1992Bonevento et al.
5175837December 1992Arnold et al.
5175839December 1992Ikeda et al.
5185864February 1993Bonevento et al.
5201053April 1993Benhase et al.
5202887April 1993Ueno et al.
5206939April 1993Yanai et al.
5210865May 1993Davis et al.
5235690August 1993Beardsley et al.
5235692August 1993Ayres et al.
5239659August 1993Rudeseal et al.
5263154November 1993Eastridge et al.
5269011December 1993Yanai et al.
5274645December 1993Idelman et al.
5276867January 1994Kenley et al.
5285451February 1994Henson et al.
5313664May 1994Sugiyama et al.
5335352August 1994Yanai et al.
5343477August 1994Yamada
5375232December 1994Legvold et al.
5377342December 1994Sakai et al.
5379412January 1995Eastridge et al.
5381539January 1995Yanai et al.
5428796June 1995Iskiyan et al.
5446872August 1995Ayres et al.
5459857October 1995Ludlam et al.
5463752October 1995Behnase et al.
5526484June 1996Casper et al.
5537533July 1996Staheli et al.
5544345August 1996Carpenter et al.
5544347August 1996Yanai et al.
5555371September 1996Duyanovich et al.
5574950November 1996Hathorn et al.
5584039December 1996Johnson et al.
5592618January 1997Micka et al.
5606359February 1997Youden et al.
5608865March 1997Midgely et al.
5611069March 1997Matola
5613155March 1997Baldiga et al.
5615329March 1997Kern et al.
5649093July 1997Hanko et al.
5673382September 1997Cannon et al.
5680580October 1997Beardsley et al.
5680640October 1997Ofek et al.
5682396October 1997Yamamoto et al.
5682513October 1997Candelaria et al.
5721916February 1998Pardikar
5724500March 1998Shinmura et al.
5742792April 1998Yanai et al.
5774668June 1998Choquier et al.
5809543September 1998Byers et al.
5815146September 1999Youden et al.
5829047October 1998Jacks et al.
5889935March 1999Ofek et al.
5901327May 1999Ofek
Foreign Patent Documents
0239323Sep., 1987EP
0323123Dec., 1988EP
1-120650May., 1989JP
1-19437Jan., 1989JP
1-19438Jan., 1989JP
2-32418Feb., 1990JP
2-32419Feb., 1990JP
2-32420Feb., 1990JP
2-91716Mar., 1990JP
2-91717Mar., 1990JP
2-93721Apr., 1990JP
2086625May., 1992GB
57-111900Jul., 1982JP
PCT/US/84/01678Jul., 1995WO
PCT/US/93/05853Jan., 1994WO
WO94/25919Nov., 1994WO
Other References
DL. Burkes & R.K. Treiber, "Design Approaches for Real-Time Transaction Processing Remote Site Recovery," Computer Society International Conference (COMPCON), Spring Meeting, Los Alamitos, Feb. 26-Mar. 2, 1990, No., Conf. 35, 23 Feb. 1990, Institute of Electrical and Electronics Engineers, New York, N.Y., pp. 568-572. .
Supplementary European Search Report for EP 94 91 4223, Jul. 1, 1997. .
International Search Report for PCT/US/94/04326, Jul. 20, 1994. .
PCT Written Opinion for PCT/US94/04326, Feb. 28, 1995. .
PCT International Preliminary Examination Report for PCT/US94/04326, Jul. 18, 1995. .
SFT Netware 286 Maintenance, #100-313-001, 100/Rev1.00, Novel Incorporated, Provo, Utah, Nov. 1987. .
SFT Netware 286 Installation, #100-312-001, 100/Rev1.00, Novel Incorporated, Provo, Utah, Nov. 1987. .
SFT Netware 286 Installation Supplement, #100-000225-001, 26/Rev1.02, Novel Incorporated, Provo, Utah, Jun. 1986. .
SFT Netware 68 Maintenance, #100-000289-001, 64/Rev1.00, Novel Incorporated, Provo, Utah, Nov. 1986. .
SFT Netware 68 Installation, #100-000317-001, 103/Rev1.00, Novel Incorporated, Provo, Utah, Feb. 1988. .
SFT Netware 868 Installation Supplement, #100-000286-001, 61/Rev1.00, Novel Incorporated, Provo, Utah, Dec. 1986. .
Symmetrix Remote Data Facility (SRDF) Product Guide, P/N 200-999-554 Rev A, EMC Corporation, Hopkinton, Mass., Sep. 1994, 105 pages [EMCP 00003686-3791]. .
"Symmetrix Remote Data Facility Data Sheet, Symmetrix-resident Software Feature to extend mirroring capabilities," EMC Corporation, Hopkinton, Mass., Apr. 1995, 4 pages [EMCP 00010741-1074]. .
"When Data is Lost in Minutes, its an Act of God. When Data is Recovered in Minutes, its an Act of EMC." EMC Corporation, Hopkinton, Mass., 1994, one page [EMCP 00003903]. .
"EMC Ships Mainframe Disaster Recovery Solution that Restores Data in Minutes," News from EMC Corporation, EMC Corporation, Hopkinton, Mass., Oct. 3, 1994, two pages [EMCP 00003921-3922]. .
"Asset Protection: Dish-Based Approach Makes Disaster Recovery More Efficient," Ensures Business Continuity, EMC Corporate Information, EMC Corporation, Hopkinton, Mass., 5 pages [EMCP 00003916-3920]. .
"EMC Ships Continuously-Available Disk-Based Disaster Recovery Solution," News from EMC Corporation, EMC Corporation, Hopkinton, Mass., Oct. 3, 1994, 4 pages [EMCP 00003923-8926]. .
"Symmetrix Remote Data Facility: Taking the disaster out of disaster recovery," Brochure, EMC Corporation, Hopkinton, Mass., Sep., 1994, 8 pages [EMCP 00003908-3915]. .
"Symmetrix Remote Data Facility: EMC redefines disaster recovery," Brochure, EMC Corporation, Hopkinton, Mass., Sep. 1994, 4 pages [EMCP 00003904-3907]. .
"EMC Introduces Symmetrix 5500 Continuous Operation Disk Storage for Mission Critical IBM Mainframe Market," News from EMC Corporation, Hopkinton, Mass., Nov. 10, 1992, 3 pages. .
"MCI Deploys New Fast Data Transfer Technology from EMC," News from EMC Corporation, Hopkinton, Mass., May 22, 1995, 2 pages [EMCP 00007520-7521]. .
"EMC Agreement with Data Switch to Promote Long Distance Disaster Recovery," News from EMC Corporation, EMC Corporation, Hopkinton, Mass., Nov. 30, 1994, 2 pages [EMCP 00007516-7517]. .
"EMC and Comdisco to Benchmark Data Mirroring Capability," News from EMC Corporation, EMC Corporation, Hopkinton, Mass., Jan. 24, 1995, 2 pages [EMCP 00007518-7519]. .
"EMC Ships Mainframe Disaster Recovery Solution that Restores Data in Minutes," News from EMC Corporation, EMC Corporation, Hopkinton, Mass., Oct. 3, 1994, 2 pages [EMCP 00007514-7515]. .
N. S. Prasad, "IBM Mainframes: Architecture and Design," McGraw-Hill Book Company, New York, N.Y., 1989, pp. 58-73 and 182-196. .
"The IBM 3390 Storage Control Reference," No. GA32-0099-04, International Business Machines Corporation, Tucson, Arizona, 1991, pp. 1-304. .
Patterson et al., "A Case for Redundant Arrays of Inexpensive Disks (RAID)," Report No. UCB/CSD 87/391, Computer Science Division (EECS), University of California, Berkeley, California, Dec. 1987, pp. 1-24. .
Patterson et al., "Introduction to Redundant Arrays of Inexpensive Disks (RAID)," COMPCON 89 Proceedings, Feb. 27--Mar. 3, 1989, IEEE Computer Society, pp. 112-117. .
Ousterhout et al., "Beating the I/O Bottleneck: A Case for Log-Structured File Systems," Operating Systems Review, vol. 23, No. 1, ACM Press, Jan., 1989, pp. 11-28. .
Douglis et al., "Log Structured File Systems," COMPCON 89 Proceedings, Feb. 27--Mar. 3, 1989, IEEE Computer Society, pp. 124-129. .
Rosemblum et al., "The Design and Implementation of a Log-Structured File System," ACM Transactions on Computer Systems, vol. 1, Feb. 1992, pp. 26-52. .
Hank Cote--New Horizons in Accessing Large Amounts of On-Line Data, pp. 71-75, 1982. .
Matt Kramer--Fault-Tolerant LANs Guard Against Malfunction, Data Loss, pp. C26-C35, Sep. 1987. .
Carol Grossman--Planning for 3990 Extended Functions, pp. 245-258, 1989, 30th Annual GUIDE Conference. .
Bob Buchanan--Administrative Error Causes Most Major System Failures (3 pgs.), Feb. 1988, Government Computer News, V. 17, N. 4, p. 340. .
Virginia Dudek--Planning for Network Disasters (4 pgs.), Jun. 12, 1989, MIS Week, vol. 10, N. 24. .
Jean Bozman--Escon An Aid To Remote Disks (3 pgs.), V. 25, N. 1, p. 29, Jan. 7, 1991. .
Gary A. Boles--A RAID Stack: Micropolis RAIDION, Network Computer, Jun. 1992. .
Blaine Homer--RAID Level 5 Protects Your Data With Interleaved Parity, LAN Times, May 25, 1992, vol. 9. .
Michael Caton--Micropolis Array Secures Data, PC Week, 1992. .
RAIDON Fault-Tolerant Disk Arrays Brochure, 1993. .
Storagetek 2Q Earnings Down, Iceberg Ready for Testing by Jim Mallory, Newsbytes, Jul. 15, 1993. .
Network Backup Evolves, by Michael Peterson et al., PC Magazine, vol. 12, No. 16, Sep. 28, 1993, p. 277(18). .
Your Next LAN Storage Could Be a Mainframe, Jeff Moad, Datamination, vol. 39, No. 6, Mar. 15, 1993, p. 71. .
Heard Any Good Rumors Lately? (Network Management), By Jill Huntington-Lee, LAN Computing, vol. 5, No. 2, Feb. 1994, p. 22. .
Special Report: The AT&T Worldwide Intelligent Network--Scope, Nature and Management, Edge, vol. 7, No. 224, Nov. 6, 1992, p. 9. .
Protect Your Data! (Database Report), by Michael Liczbanski, Data Based Advisor, vol. 10, No. 5, May 1992, p. 114. .
Network Planning to the Rescue; Network Managers Should Design and Test a Network Recovery Plan That Lets Business Continue Uninterrupted Even If Disaster Strikes, by Ramond W. Rudnitskas et al., Networking Management, vol. 9, No. 11, Oct. 1991, p. 46. .
If LANS Eye Host Status, Quality Today Falls Short, by Barbara Bochenski, Software Magazine, vol. 11, No. 12, Oct. 1991, p. 66. .
Safe and Secure, by Peter Stephenson, LAN Magazine, vol. 6, No. 9, Sep. 1991, p. 34. .
Firing Up Fault Tolerance by Brian O'Connell, DEC Professional, vol. 10, No. 6, Jun. 1991, p. 118. .
A Which Computer? Guide to Backup Systems, by Chris Bidmead, Which Computer?, Sep. 1989, p. 62. .
Optical Disks Compete With Videotape and Magnetic Storage Media, by Henry Urrows et al., Optical Information Systems, vol. 8, No. 3, May-Jun. 1988, p. 101. .
Preventing Disasters: Whipping Up an Emergency Response Plan, Network Computing, Feb. 1992, p. 81. .
Abbey Banking on FDDI, Communications Week International, Mar. 4, 1991, p. 22. .
X.25 Data Networks and IBM: A Troublesome Connection, by Jan Johnson, Computer Decisions, vol. 19, Jan. 12, 1987, p. 16. .
DSC Communications Corp. Debuts NEXOS LAN Product Line at Interface, Demos Enhanced Performance, Fault Tolerance and WAN Connectivity, News Release, Mar. 28, 1988. .
Analysis of Scanning Policies For Reducing Disk Seek Times, by E. G. Coffman, L. A. Klimko and Barbara Ryan-Siam, Journal on Computing, Sep. 1972, vol. 1, No. 3, pp. 269-280. .
Configuring Tandem Disk Subsystems, by Scott Sitler, Tandem Systems Review, Dec. 1986, pp. 80-91. .
Technical Report Investigation of Selected Disk Systems, Oct. 1976, Teledyne Brown Engineering, Huntsville, Alabama (18 pages). .
Software Extends Unix Distributed Computing, by Evan O. Grossman, PC Week, Jun. 5, 1989, p. 35 and p. 40. .
IBM Funds Pack Firm, Electronic News, May 22, 1989, pp. 12. .
IBM Plans U.S. Test of File-Transfer, Network World, Sep. 5, 1988, pp. 15, 17. .
IBM Invests In Firm That Offers Unix File-Sharing, Network World, May 22, 1989, p. 4. .
IBM Plans Speedy Nationwide File System, Computer Systems News, Sep. 5, 1988, p. 56. .
Design Alternatives for Disk Duplexing, by Spencer W. Ng, IBM Research Report, Jan. 30, 1987. .
Disk Arm Movement In Anticipitation of Future Requests, by Richard P. King, IBM Research Report, Dec. 16, 1987. .
Ubik: Replicated Servers Made Easy, by Michael Leon Kazar, pp. 60-67, 1989. .
Carnegie-Mellon Forges Standard Look For Multivendor Net, Computerworld, Jan. 30, 1989, pp. SR/8-SR/9. .
A Multi-Media Message System For Andrew, USENIX Winter Conference, Feb. 9-12, 1988, pp. 37-42. .
Synchronization and Caching Issues in the Andrew File System, USENIX, Winter Conference, Feb. 9-12, 1988, pp. 27-36. .
Comparative Study of Some UNIX Distributed File Systems, EUUG, Sep. 22-24, 1986, pp. 73-82. .
VOLUMES-The Andrew File System Data Structuring Primitive, EUUG Autumn 1986, Sep. 22-24, 1986, pp. 473-480. .
Unix Plays Role in Corporate Networking, MIS Week, vol. 10, No. 28, Jul. 17, 1989, pp. 22-23. .
Sun's Open System Made by Marketplace, "Unlike Minis, Unix Machines, Pcs and Even VMS Can Be Glued" by NFS, Software Magazine, vol. 9, No. 12, Oct. 1989, pp. 72-75, Cashin, Jerry. .
Universal Server Under Development, MacWEEK, vol. 3, No. 39, Oct. 31, 1989, p. 5. .
Unix Plays Role In Corporate Networking-Vendors Look Unix-Based Network Management Tools as Networking Standards Slowly Evolve, MIS Week, Jul. 17, 1989, pp. 22-23, vol. 10, No. 28. .
The 28 Runners Declared for Open Software Foundations, Distributed Computing Stakes, Computergram International, No. 1301, CGI11909912, Nov. 9, 1989. .
Some Platforms (Theory of Self-Organizing Systems), Release 1.0, vol. 89, No. 6, Jun. 27, 1989, pp. 12-16. .
Self-Organizing Systems (Overview of Detailed Examples of Self-Organizing Systems, Includes related article About the Importance of Self-Organizing Systems), Release 1.0, vol. 89, No. 6, Jun. 27, 1989, pp. 1-9. .
IBM-Funded Startup Developing Unix Add-On, PC Week, vol. 6, No. 22, Jun. 5, 1989, p. 40. .
Uniting File Systems; Experimental Large Scale, Distributed File Systems Are Now Being Created (included related article on the History of the AFS Project), UNIX Review, vol. 7, No. 3, Mar. 1989, pp. 61-70. .
IBM Backs Transarc, Formed to Offer Andrew File System Distributed Unix Database, Computergram International, No. 1180, CGI05190001, May 19, 1989. .
A Comparative Analysis of Disk Scheduling Policies, by Toby J. Reorey and Tad B. Pinerton, Communications of the ACM, Mar. 1972, vol. 15, No. 3, pp. 177-184. .
Disk Scheduling: FCFS vs. SSTF Revisited, by Micha Hofri, Communications of the ACM, Nov. 1980, vol. 23, No. 11, pp. 645-653. .
Using Multiple Replica Classes to Improve Performance in Distributed System, by Peter Triantafillou and David Tayloer, The 11th International Conference on Distributed Computing Systems, May 20-24, 1991, pp. 420-428. .
Government Computer News, Oct. 9, 1987, Richard A. Damca, Microcomputing "Patch Fault Tolerance: A Strong Second Team," vol. 7, Issue 149 P27(2), Dec. 17, 1990. .
Computer Reseller, Feb. 19, 1990, Joel Shore, Mirrored Servers' a reality, pp. 51, 52, 62. .
Infoworld Articles, Mar. 12, 1990, Vendors Offer Mirroring Solutions P(31). .
Wilkinson, Compact and Novell team for fault tolerance: mirrored servers to cut down on costly downtime, Article MIS Week, vol. 11, Issue 7, Feb. 12, 1990. .
Article PC Week, vol. 4, Issue 37, Sep. 15, 1987, Matt Kramer, Fault-Tolerant LANS Guard Against Malfunction, Data Loss. .
"The DASD Time Bomb," Datamation, vol. 32, Mar. 1, 1986, P63(4), Alper, Alan..~
Primary Examiner: Thai; Tuan V.
Attorney, Agent or Firm:Arnold White & Durkee
Parent Case Text


RELATED APPLICATION

This Application is a divisional of U.S. patent application Ser. No. 08/654,511 filed May 28, 1996, entitled REMOTE DATA MIRRORING by Yanai et al, issued as U.S. Pat. No. 5,742,792 on Apr. 21, 1998.
Claims

What is claimed is:
1. A data processing system comprising:
a first data storage system;
a second data storage system; and
a data link coupled between said first data storage system and said second data storage system for transmission of remote copy data from said first data storage system to said second data storage system;
wherein said first data storage system is operational in a remote copy mode in which data is stored in a primary volume of data storage in said first data storage system and is transmitted over said data link and stored in a respective secondary volume of data storage in said second data storage system; and
said first data storage system includes a controller for detecting an occurrence of a disruption in operation of the data processing system preventing an access to said primary volume, and upon detecting said occurrence of said disruption in operation of the data processing system preventing said access to said primary volume, checking whether an automatic recovery mode has been preselected for said primary volume, and
when said automatic recovery mode has been preselected for said primary volume, automatically accessing said secondary volume in lieu of said primary volume, and
when said automatic recovery mode has not been preselected for said primary volume, not automatically accessing said secondary volume and instead signaling that intervention is required.

2. The data processing system as claimed in claim 1, wherein the data storage in the first data storage system and the data storage in the second data storage system are configured to define volume pairs, each volume pair including a primary volume in one of the first data storage system and the second data storage system and a corresponding secondary volume in another of the first data storage system and the second data storage system, said volume pairs including primary volumes each having an attribute for selecting or not selecting said automatic recovery mode.

3. The data processing system as claimed in claim 1, wherein said controller has control logic for detecting an occurrence of a disruption in operation of the data processing system preventing an access to said secondary volume, and upon detecting said occurrence of said disruption in operation of the data processing system preventing said access to said secondary volume, checking whether said automatic recovery mode has been preselected for said secondary volume, and
when said automatic recovery mode has been preselected for said secondary volume, automatically accessing said primary volume in lieu of said secondary volume, and
when said automatic recovery mode has not been preselected for said secondary volume, not automatically accessing said primary volume and instead signaling that intervention is required.

4. The data processing system as claimed in claim 1, wherein the data storage in said first data storage system and in said second data storage system are configured to define volume pairs, each volume pair including a primary volume in one of the first data storage system and the second data storage system and a corresponding secondary volume in another of the first data storage system and the second data storage system, each volume pair having an attribute for selecting or not selecting said automatic recovery mode with respect to the primary volume and the secondary volume in said each volume pair.

5. The data processing system as claimed in claim 1, wherein the controller includes control logic for sending an "intervention required" message to an operating system of a host computer in order to signal that intervention is required.

6. A data storage system comprising:
primary data storage; and
a data storage controller programmed to operate in a remote copy mode in which data is stored in a primary volume of the primary data storage and is transmitted over a data link to remote data storage for storage in a respective secondary volume of the remote data storage;
wherein said data storage controller is further programmed to detect an occurrence of a disruption in operation of the data storage system preventing an access to said primary volume, and upon detecting said occurrence of said disruption in operation of the data storage system preventing said access to said primary volume, to check whether an automatic recovery mode has been preselected for said primary volume, and
when said automatic recovery mode has been preselected for said primary volume, to automatically access said secondary volume in lieu of said primary volume, and
when said automatic recovery mode has not been preselected for said primary volume, to not automatically access said secondary volume and instead to signal that intervention is required.

7. The data storage system of claim 6, wherein said data storage controller is programmed to detect an occurrence of a disruption in operation of the data storage system preventing an access to said secondary volume, and upon detecting said occurrence of said disruption in operation of the data storage system preventing said access to said secondary volume, to check whether said automatic recovery mode has been preselected for said secondary volume, and
when said automatic recovery mode has been preselected for said secondary volume, to automatically access said primary volume in lieu of said secondary volume, and
when said automatic recovery mode has not been preselected for said secondary volume, to not automatically access said primary volume and instead to signal that intervention is required.

8. A program storage device readable by a data storage system, said program storage device encoding a program for execution by the data storage system for controlling operation of the data storage system in a remote copy mode in which data is stored in a primary volume of primary data storage of the data storage system and is transmitted over a data link to remote data storage for storage in a respective secondary volume of the remote data storage;
wherein said program is executable by the data storage system to detect an occurrence of a disruption in operation of the data storage system preventing an access to said primary volume, and upon detecting said occurrence of said disruption in operation of the data storage system preventing said access to said primary volume, to check whether an automatic recovery mode has been preselected for said primary volume, and
when said automatic recovery mode has been preselected for said primary volume, to automatically access said secondary volume in lieu of said primary volume, and
when said automatic recovery mode has not been preselected for said primary volume, to not automatically access said secondary volume and instead to signal that intervention is required.

9. The program storage device as claimed in claim 8, wherein said program is executable by the data storage system to detect an occurrence of a disruption in operation of the data storage system preventing an access to said secondary volume, and upon detecting said occurrence of said disruption in operation of the data storage system preventing said access to said secondary volume, to check whether said automatic recovery mode has been preselected for said secondary volume, and
when said automatic recovery mode has been preselected for said secondary volume, to automatically access said primary volume in lieu of said secondary volume, and
when said automatic recovery mode has not been preselected for said secondary volume, to not automatically access said primary volume and instead to signal that intervention is required.

10. A method of recovering from disruptions in a data processing system, said data processing system having a first data storage system, a second data storage system, and a data link coupled between said first data storage system and said second data storage system for transmission of remote copy data from said first data storage system to said second data storage system; said method comprising the steps of:
(a) configuring the data storage in said first data storage system and in said second data storage system to define volume pairs each including a primary volume in said first data storage system and a corresponding secondary volume in the second data storage system, said volume pairs including primary volumes each having an attribute for selecting or not selecting an automatic recovery mode;
(b) operating said data processing system in a remote copy mode in which data is stored in the primary volumes of data storage in said first data storage system and is transmitted over said data link and stored in the corresponding secondary volumes of data storage in said second data storage system;
(c) detecting an occurrence of a disruption in operation of the data processing system preventing an access to a primary volume having said attribute selecting said automatic recovery mode, and upon detecting said occurrence of said disruption in operation of the data processing system preventing said access to the primary volume having said attribute selecting said automatic recovery mode, automatically accessing the secondary volume in the volume pair including the primary volume having said attribute selecting said automatic recovery mode; and
(d) detecting an occurrence of a disruption in operation of the data processing system preventing an access to a primary volume having said attribute not selecting said automatic recovery mode, and upon detecting said occurrence of said disruption in operation of the data processing system preventing said access to the primary volume having said attribute not selecting said automatic recovery mode, not automatically accessing the secondary volume in the volume pair including the primary volume having said attribute not selecting said automatic recovery mode, and instead signaling that intervention is required.

11. The method as claimed in claim 10, wherein said volume pairs include secondary volumes each having an attribute for selecting or not selecting said automatic recovery mode, and wherein said method further includes the steps of:
(e) detecting an occurrence of a disruption in operation of the data processing system preventing an access to a secondary volume having said attribute selecting said automatic recovery mode, and upon detecting said occurrence of said disruption in operation of the data processing system preventing said access to the secondary volume having said attribute selecting said automatic recovery mode, automatically accessing the primary volume in the volume pair including the secondary volume having said attribute selecting said automatic recovery mode; and
(f) detecting an occurrence of a disruption in operation of the data processing system preventing an access to a secondary volume having said attribute not selecting said automatic recovery mode, and upon detecting said occurrence of said disruption in operation of the data processing system preventing said access to the secondary volume having said attribute rot selecting said automatic recovery mode, not automatically accessing the primary volume in the volume pair including the secondary volume having said attribute not selecting said automatic recovery mode, and instead signaling that intervention is required.

12. The method as claimed in claim 10, wherein the first data storage system sends an "intervention required" message to an operating system of a host computer in order to signal that intervention is required.

13. A data processing system comprising:
a first data storage system;
a second data storage system; and
a data link coupled between said first data storage system and said second data storage system for transmission of data from said first data storage system to said second data storage system;
wherein said first data storage system is operational in a remote copy mode in which remote copy data is stored in data storage in said first data storage system and is transmitted over said data link and stored in data storage in said second data storage system; and
said first data storage system includes a controller for detecting a failure of said data link preventing the remote copy data from being transmitted to said second data storage system for storage in the data storage in said second data storage system, and upon detecting said failure of said data link preventing the remote copy data from being transmitted to said second data storage system for storage in the data storage in said second data storage system, checking whether an automatic recovery mode has been preselected for use in case of link failure, and
when said automatic recovery mode has been preselected for use in case of link failure, permitting continued access to the remote copy data in the data storage in the first data storage system without intervention, and
when said automatic recovery mode has not been preselected for use in case of link failure, not permitting continued access to the remote copy data in the data storage of the first data storage system without intervention and instead signaling that intervention is required.

14. The data processing system as claimed in claim 13, wherein the controller includes control logic for sending an "intervention required" message to an operating system of a host computer in order to signal that intervention is required.

15. A data storage system comprising:
primary data storage; and
a data storage controller programmed to operate in a remote copy mode in which remote copy data is stored in the primary data storage and is transmitted over a data link to remote data storage;
wherein said data storage controller is further programmed to detect a failure of said data link preventing the remote copy data from being transmitted to said remote data storage, and upon detecting said failure of said data link preventing the remote copy data from being transmitted to said remote data storage, to check whether an automatic recovery mode has been preselected for use in case of link failure, and
when said automatic recovery mode has been preselected for use in case of link failure, to permit continued access to the remote copy data in the primary data storage without intervention, and
when said automatic recovery mode has not been preselected for use in case of link failure, to not permit continued access to the remote copy data in the primary data storage without intervention and instead to signal that intervention is required.

16. A program storage device readable by a data storage system, said program storage device encoding a program for execution by the data storage system for controlling operation in a remote copy mode in which remote copy data is stored in primary data storage of the data storage system and is transmitted over a data link to remote data storage, wherein said program s executable by the data storage system to detect a failure of said data link preventing the remote copy data from being transmitted to said remote data storage, and upon detecting said failure of said data link preventing the remote copy data from being transmitted to said remote data storage, to check whether an automatic recovery mode has been preselected for use in case of link failure, and
when said automatic recovery mode has been preselected for use in case of link failure, to permit continued access to the remote copy data without intervention, and
when said automatic recovery mode has not been preselected for use in case of link failure, to not permit continued access to the remote copy data without intervention and instead to signal that intervention is required.

17. A method of operating a data processing system, said data processing system including a first data storage system, a second data storage system, a data link coupled between said first data storage system and said second data storage system for transmission of remote copy data from said first data storage system to said second data storage system, and a host computer coupled to said first data storage system for access to data storage in said first data storage system; said method comprising the steps of:
(a) selecting whether or not an automatic recovery mode should be used in case of link failure; and then
(b) operating said data processing system in a remote copy mode in which remote copy data is stored in the data storage in said first data storage system and is transmitted over said data link and stored in data storage in said second data storage system; and
(c) detecting a failure of said data link preventing data stored in the data storage in said first data storage system from being transmitted to said second data storage system for storage in the data storage in said second data storage system, and upon detecting said failure of said data link preventing remote copy data stored in the data storage in said first data storage system from being transmitted to said second data storage system for storage in the data storage in said second data storage system, checking whether said automatic recovery mode has been preselected for use in case of link failure, and
when said automatic recovery mode has been preselected for use in case of link failure, permitting said host computer to continue to access the remote copy data in the data storage in the first data storage system without intervention, and
when said automatic recovery mode has not been preselected for use in case of link failure, not permitting said host computer to access the remote copy data in the data storage in the first data storage system without intervention and instead signaling that intervention is required.

18. The method as claimed in claim 17, wherein the first data storage system includes means for sending an "intervention required" message to an operating system of said host computer in order to signal that intervention is required.

19. The method as claimed in claim 17, wherein said automatic recovery mode has not been preselected for use in case of link failure, said failure of the data link occurs preventing data stored in the data storage in said first data storage system from being transmitted to said second data storage system for storage in the data storage in said second data storage system, and the host computer is prevented from accessing the remote copy data in the data storage of the first data storage system for read access and write access.
Description
FIELD OF THE INVENTION

This invention relates to data storage, and more particularly, to a system and method for automatically providing and maintaining a copy or mirror of data stored at a location remote from the main or primary data storage device.

BACKGROUND OF THE INVENTION

Nearly all data processing system users are concerned with maintaining back-up data in order to insure continued data processing operations should their data become lost, damaged, or otherwise unavailable.

Large institutional users of data processing systems which maintain large volumes of data such as banks, insurance companies, and stock market traders must and do take tremendous steps to insure back up data availability in case of a major disaster. These institutions recently have developed a heightened awareness of the importance of data recovery and back-up in view of the many natural disasters and other world events including the bombing of the World Trade Center in New York City.

Currently, data processing system users often maintain copies of their valuable data on site on either removable storage media, or in a secondary "mirrored" storage device located on or within the same physical confines of the main storage device. Should a disaster such as fire, flood, or inaccessibility to a building occur, however, both the primary as well as the secondary or backed up data will be unavailable to the user. Accordingly, more data processing system users are requiring the remote storage of back up data.

One prior art approach at data back-up involves taking the processor out of service while back-up tapes are made. These tapes are then carried off premises for storage purposes. Should access to the backed up data be required, the proper tape must be located, loaded onto a tape drive, and restored to the host system requiring access to the data. This process is very time consuming and cost intensive, both in maintaining an accurate catalog of the data stored on each individual tape, as well as storing the large number of tapes required to store the large amounts of data required by these institutions. Additionally and most importantly, it often takes twenty-four hours before a back-up tape reaches its storage destination during which time the back-up data is unavailable to the user.

Additionally, today's systems require a significant amount of planning and testing in order to design a data recovery procedure and assign data recovery responsibilities. Typically, a disaster recovery team must travel to the test site carrying a large number of data tapes. The team then loads the data onto disks, makes the required network connections, and then restores the data to the "test" point of failure so processing can begin. Such testing may take days or even weeks and always involves significant human resources in a disaster recovery center or back-up site.

Some providers of prior art data storage systems have proposed a method of data mirroring whereby one host Central Processing Unit (CPU) or processor writes data to both a primary, as well as a secondary, data storage device or system. Such a proposed method, however, overly burdens the host CPU with the task of writing the data to a secondary storage system and thus dramatically impacts and reduces system performance.

Accordingly, what is required is a data processing system which automatically and asynchronously, with respect to a first host system, generates and maintains a back-up or "mirrored" copy of a primary storage device at a location physically remote from the primary storage device, without intervention from the host which seriously degrades the performance of the data transfer link between the primary host computer and the primary storage device.

SUMMARY OF THE INVENTION

This invention features a system which controls storing of primary data received from a primary host computer on a primary data storage system, and additionally controls the copying of the primary data to a secondary data storage system controller which forms part of a secondary data storage system, for providing a back-up copy of the primary data on the secondary data storage system which is located in a remote location from the primary data storage system. For remote copying of data from one storage system to the other without host involvement, the primary and secondary data storage system controllers are coupled via at least one high speed communication link such as a fiber optic link driven by LED's or laser. The high speed communication link also permits one data storage system to read or write data to or from the other data storage system.

At least ore of the primary and secondary data storage system controllers coordinates the copying of primary data to the secondary data storage system and at least one of the primary and secondary data Storage system controllers maintains at least a list of primary data which is to be copied to the secondary data storage device.

Additionally, the secondary data storage system controller provides an indication or acknowledgement to the primary data storage system controller that the primary data to be copied to the secondary data storage system in identical form as secondary data has been received or, in another embodiment, has actually been written to a secondary data storage device.

Accordingly, data may be transferred between the primary and secondary data storage system controllers synchronously, when a primary host computer requests writing of data to a primary data storage device, or asynchronously with the primary host computer requesting the writing of data to the primary data storage system, in which case the remote data copying or mirroring is completely independent of and transparent to the host computer system.

At least one of the primary data storage system controller and the secondary data storage system controller maintains a list of primary data which is to be written to the secondary data storage system. Once the primary data has been at least received or optionally stored on the secondary data storage system, the secondary data storage system controller provides an indication or acknowledgement of receipt or completed write operation to the primary data storage system.

At such time, the primary and/or secondary data storage system controller maintaining the list of primary data to be copied updates this list to reflect that the given primary data has been received by and/or copied to the secondary data storage system. The primary or secondary data storage system controllers and/or the primary and secondary data storage devices may also maintain additional lists for use in concluding which individual storage locations, such as tracks on a disk drive, are invalid on any given data storage device, which data storage locations are pending a format operation, which data storage device is ready to receive data, and whether or not any of the primary or secondary data storage devices are disabled for write operations.

The remote mirroring facility can operate in a specified one of a number of different remote mirroring operating modes for each volume. The operating modes include a synchronous mode, a semi-synchronous mode, an adaptive copy--write pending mode, and an adaptive copy-disk mode. The operating mode for each logical volume can be specified to best suit the purposes of the desired remote mirroring, the particular application using the volume, and the particular use of the data stored on the volume.

In the synchronous mode, data on the primary (R1) and secondary (R2) volumes are always fully synchronized at the completion of an I/O sequence. The data storage system containing the primary (R1) volume informs the host that an I/O sequence has successfully completed only after the data storage system containing the secondary (R2) volume acknowledges that it has received and checked the data. All accesses (reads and writes) to the remotely mirrored volume to which a write has been performed are suspended until the write to the secondary (R2) volume has been acknowledged.

In the semi-synchronous mode, the remotely mirrored volumes (R1, R2) are always synchronized between the primary (R1) and the secondary (R2) prior to initiating the next write operation to these volumes. The data storage system containing the primary (R1) volume informs the host that an I/O sequence has successfully completed without waiting for the data storage system containing the secondary (R2) volume to acknowledge that it has received and checked the data. Thus, a single secondary (R2) volume may lag its respective primary volume (R1) by only one write. Read access to the volume to which a write has been performed is allowed (steps 405, 403) while the write is in transit to the data storage system containing the secondary (R2) volume.

The adaptive copy modes transfer data from the primary (R1) volume to the secondary (R2) volume and do not wait for receipt acknowledgment or synchronization to occur. The adaptive copy modes are responsive to a user-configurable skew parameter specifying a maximum allowable write pending tracks. When the maximum allowable write pending tracks is reached, then write operations are suspended, and in a preferred arrangement, write operations are suspended by defaulting to a predetermined one of the synchronous Dr asynchronous modes. In the adaptive copy--write pending mode, the write pending tracks accumulate in cache. In the adaptive copy--disk mode, the write pending tracks accumulate in disk memory.

In accordance with a basic aspect of the invention, there are provided a number of automatic and non-automatic recovery mechanisms. The recovery mechanism can be also selected on a logical volume basis for a desired level of data integrity and degree of operator or application program involvement. The invention also provides various options that provide a tradeoff between the degree of data integrity, cache loading, processing speed, and link traffic.

In one embodiment, cache loading and processing speed is enhanced by queuing pointers to data in cache for transmission to the link, and permitting pending write data to be overwritten in cache. Link traffic can also be reduced in this case, since obsolete write pending data need not be transmitted over the link. However, unless the remote mirroring is operated in the synchronous mode, data integrity is subject to the possibility of a "rolling disaster." In the rolling disaster, a remote mirroring relationship exists between the two data storage systems. All links break between the sites, and application processing continues using the primary (R1) volumes. The links are restored, and resynchronization commences by copying data from the primary (R1) volumes to the secondary (R2) volumes. Before resynchronization is finished, however, the primary volumes are destroyed, and the attempt at resynchronization has further corrupted the secondary volumes, due to the cache overwrite option.

The invention provides options other than the synchronous and semi-synchronous operating modes to avoid the "rolling disaster" possibility when performing automatic recovery. One option is to suspend processing whenever the host requests a write to write-pending data in cache. Another option is to log multiple versions of tracks containing remote pending data.

Another aspect of the present invention provides mechanisms for selectively inhibiting automatic or manual recovery when automatic or manual recovery would be inappropriate.

In accordance with another aspect of the invention, automatic recovery is selectively inhibited by domino modes. If a "volume domino mode" is enabled for a remotely mirrored volume pair, access to a volume of the remotely mirrored volume pair is denied when the other volume is inaccessible. In a "links domino mode," access to all remotely mirrored volumes is denied when remote mirroring is disrupted by an all-links failure.

The domino modes can be used to initiate application-based recovery in lieu of automatic recovery. In one application-based recovery scheme, an application program maintains a log file of all writes ("before" or "after" images) to a data file. To ensure recovery, the application program always writes data to the primary (R1) copy of the log file before it is written to the primary (R1) copy of the data file. The degree of synchronization between the secondary (R2) and primary (R1) copies is selected so that the remote mirroring facility always writes data to the secondary (R2) copy of the log file before it is written to the secondary (R2) copy of the data file. Therefore, in the case of an all-links failure where host processing continues so far beyond the failure that all versions of the following updates are not retained, the secondary (R2) copy of the data file can be recovered if the primary (R1) copies are destroyed. In this case, if the secondary (R2) copy of the data file is corrupted, it is recovered using the changes recorded in the secondary (R2) copy of the log file.

In accordance with yet another aspect of the invention, there is provided host remote mirroring software for permitting a system operator or host application program to monitor and control remote mirroring, migration, and recovery operations. The host remote mirroring software provides the capability of changing the configuration of the remotely mirrored volumes in the data processing system, suspending and resuming remote mirroring for specified remotely mirrored volume pairs, synchronizing specified remotely mirrored volume pairs and notifying the system operator or host application program when synchronization is achieved, invalidating or validating specified remotely mirrored volume pairs, and controlling or limiting the direction of data transfer between the volumes in a specified remotely mirrored pair.

The present invention therefore provides a data storage system which achieves nearly 100 percent data integrity by assuring that all data is copied to a remote site, and in those cases when a back-up copy is not made due to an error of any sort, an indication is stored that the data has not been copied, but instead must be updated at a future time. The system operator or application programmer is free to choose a variety of remote mirroring and recovery operations best suited for a desired processing speed and level of data integrity.

Such a system is provided which is generally lower in cost and requires substantially less manpower and facilities to achieve than the prior art devices.

BRIEF DESCRIPTION OF THE DRAWINGS

These and other features and advantages of the present invention will be better understood when read together with the following drawings wherein:

FIG. 1 is a block diagram illustrating the system with remote data mirroring according to the present invention;

FIG. 2 is a schematic representation of a portion of an index or list maintained by the system of the present invention to determine various features including which primary data has been copied to a secondary disk;

FIG. 3 is a schematic representation of an additional list or index maintained by the system of the present invention to keep track of additional items including an invalid data storage device track, device ready status and write disable device status;

FIG. 4 is a block diagram showing a preferred construction for the remotely mirrored primary and secondary data storage systems and links;

FIG. 5 is a block diagram of a short distance option for linking two geographically separated data storage systems;

FIG. 6 is a block diagram of a long distance option for linking two geographically separated data storage systems;

FIG. 7 is a first portion of a flowchart showing the operation of a channel adapter when providing data access in the synchronous and semi-synchronous remote mirroring modes;

FIG. 8 is a second portion of the flowchart showing the operation of a channel adapter when providing data access in the synchronous and semi-synchronous remote mirroring modes;

FIG. 9 is a flowchart showing a modification of FIG. 7 for adaptive copy remote mirroring modes;

FIG. 10 is a flowchart showing operation of a data storage system when a host requests a state change to a secondary (R2) volume in the data storage system;

FIG. 11 is a flowchart showing operation of a channel adapter when responding to various failures depending on whether or not an "all-links domino mode" or a "volume domino mode" is enabled;

FIG. 12 is a block diagram illustrating the use of an application-based recovery program in a data processing system employing remotely-mirrored data storage systems;

FIG. 13 is a flowchart showing the invocation and execution of the application-based recovery program for the data processing system of FIG. 12;

FIG. 14 is a first portion of a flowchart showing an iterative routine for migrating a volume concurrent with host access to the volume;

FIG. 15 is a second portion of the flowchart begun in FIG. 14;

FIG. 16 is a flowchart showing how a channel adapter maintains remote write pending bits, remote invalid bits, and remote invalid track counts in the data processing system of FIG. 4;

FIG. 17 is a flowchart showing an iterative routine using the remote write pending bits, remote invalid bits, and remote invalid track counts for migrating a volume concurrent with host access to the volume;

FIG. 18 is a block diagram showing data structures in the cache memory of the data processing system of FIG. 4;

FIG. 19 is a first portion of a flowchart showing how a host processor bundles remote write commands from all of the channel command words (CCW) in a single CCW chain into a single write command transmitted over a link to a remote data storage system;

FIG. 20 is a second portion of the flowchart begun in FIG. 19;

FIG. 21 a flowchart showing the operation of a link adapter in the data processing system of FIG. 4;

FIG. 22 is a first portion of a flowchart of the operation of a channel adapter when writing a record to a primary (R1) volume located in the same data storage system containing the primary (R1) volume; and

FIG. 23 is a second portion of the flowchart begun in FIG. 22.

While the invention is susceptible to various modifications and alternative forms, specific embodiments thereof have been shown by way of example in the drawings and will be described in detail herein. It should be understood, however, that it is not intended to limit the invention to the particular forms disclosed, but to the contrary, the intention is to cover all modifications, equivalents, and alternatives falling within the scope of the invention as defined by the appended claims.

DETAILED DESCRIPTION OF THE INVENTION

A. Overview

The present invention features a system which provides a remote mirrored data storage system which contains generally identical information to that stored on a primary data storage system. Utilizing such a system, data recovery after a disaster can be nearly instantaneous and may require little, if any, human intervention. Using the present system, the data is retrieved from a remote device through the host data processing system.

A system in accordance with the present invention is shown generally at 10, FIG. 1, and includes at site A, which is a first geographic location, a host computer system 12 as is well known to those skilled in the art. The host computer system 12
is coupled to a first and primary data storage system 14. The host 12 writes data to and reads data from the primary data storage system 14.

The primary data storage system 14 includes a primary data storage system controller 16 which receives data from the host 12 over data signal path 18. The primary data storage system controller 16 is also coupled to a storage device 20 which may include a plurality of data storage devices 22a-22c. The storage devices may include disk drives, optical disks, CD's or other data storage devices. The primary system controller 16 is coupled to the storage device 20 by means of data signal path 24.

The primary data storage system controller 16 includes at least one channel adapter (C.A.) 26 which is well known to those skilled in the art and interfaces with host processing system 12. Data received from the host is typically stored in cache
28 before being transferred through disk adapter (D.A.) 30 over data signal path 24 to the primary storage device 20. The primary data storage controller 16 also includes a data director 32 which executes one or more sets of predetermined micro-code to control data transfer between the host 12, cache memory 28, and the storage device 20. Although the data director 32 is shown as a separate unit, either one of a channel adapter 26 or disk adapter 30 may be operative as a data director, to control the operation of a given data storage system controller. Such a reconfigurable channel adapter and disk adapter is disclosed in U.S. Pat. No. 5,335,352 entitled RECONFIGURABLE, MULTI-FUNCTION DATA STORAGE SYSTEM CONTROLLER SELECTIVELY OPERABLE AS AN INPUT CHANNEL ADAPTER AND A DATA STORAGE UNIT ADAPTER, which is fully incorporated herein by reference.

The primary data storage system 14 according to one embodiment of the present invention also includes a service processor 34 coupled to the primary data storage system controller 16, and which provides additional features such as monitoring, repair, service, or status access to the storage system controller 16.

The primary data storage system controller 16 of the present invention also features at least a second disk adapter 36 coupled to the internal bus 38 of the primary data processing system controller 16. The second disk adapter 36 is coupled, via a high speed communication link 40 to a disk adapter 42 on a secondary data storage system controller 44 of a secondary data storage system 46. Such high speed, point-to-point communication links between the primary and secondary data processing system controllers 16 and 44 include a fiber optic link driven by an LED driver, per IBM ESCON standard; a fiber optic link driven by a laser driver, and optionally T1 and T3 telecommunication links. Utilizing network connections, the primary and secondary data storage system controllers 16 and 44 may be connected to FDDI networks, T1 or T3 based networks and SONET networks.

The secondary data storage system 46 is located at a second site geographically removed from the first site. For this patent application, "geographically removed site" means not within the same building as the primary data storage system. There are presently known data processing systems which provide data mirroring to physically different data storage systems. The systems, however, are generally within the same building. The present invention is directed to providing complete data recovery in case of disaster, such as when a natural disaster such as a flood or a hurricane, or man made disasters such as fires or bombings destroy one physical location, such as one building.

As in the case of the primary data storage system, the secondary data storage system 46 includes, in addition to the secondary data storage system controller 44, a secondary data storage device 48 including a plurality of storage devices 50a-50c. The plurality of storage devices on the secondary data storage system 46, as well as the primary data storage system 14, may have various volumes and usages such as a primary data storage device 50a which is primary with respect to the attached storage controller 44 and host 52 in the case of the secondary data storage system 46, and the primary storage device 22a which is primary with respect to the first or primary host 12 in the case of the primary data storage system 14.

Additionally, each storage device, such as storage device 48, may include a secondary storage volume 50b which serves as the secondary storage for the primary data stored on the primary volume 22a of the primary data storage system 14. Similarly, the primary data storage system 14 may include a secondary storage volume 22b which stores primary data received and copied from the secondary site and data processing system 46 and host 52.

Additionally, each storage device 20, 48, may include one or more local volumes or storage devices 22c, 50c, which are accessed only by their locally connected data processing systems.

The secondary storage system controller 44 also includes at least a first channel adapter 54 which may receive data from an optionally connected secondary host 52 or an optionally connected hotsite host or CPU 56. Optionally, the primary host 12
may include a data signal path 58 directly into the channel adapter 54 of the secondary data storage system 46, while the optional secondary host 52 may include an optional data path 60 into the channel adapter 26 of the primary data storage system 14. Although the secondary host 52 illustrated in FIG. 1 is not required for remote data mirroring as described in the present patent application, such a host would be required for data retrieval if both the primary host 12 as well as the primary data storage system 14 would be rendered inoperative. Similarly, a hotsite host or CPU 56 may optionally be provided at a third geographically remote site to access the data stored in the secondary data storage system 46.

The high speed link 40 between the primary and secondary data storage systems 14 and 46 is designed such that multiple links between the primary and secondary storage system may be maintained for enhanced availability of data and increased system performance. The number of links is variable and may be field upgradeable. Additionally, the service processor 34 of the primary data storage system 14 and the service processor 62 of the secondary data storage system 46 may also be coupled to provide for remote system configuration, remote software programming, and a host base point of control of the secondary data storage system.

The secondary data storage system controller 44 also includes cache memory 64 which receives data from channel adapter 54 and disk adapter 42, as well as disk adapter 66 which controls writing data to and from secondary storage device 48. Also provided is a data director 68 which controls data transfer over communication bus 70 to which all the elements of the secondary data storage system controller are coupled.

An additional feature of the system of FIG. 1 is the ability to dynamically reconfigure channel adapters as disk adapters and disk adapters as channel adapters, as described in U.S. Pat. No. 5,269,011 entitled DYNAMICALLY RECONFIGURABLE DATA STORAGE SYSTEM WITH STORAGE SYSTEM CONTROLLERS SELECTIVELY OPERABLE AS CHANNEL ADAPTERS OR STORAGE DEVICE ADAPTERS which is fully incorporated herein by reference.

The primary and secondary data storage systems may optionally be connected by means of currently available, off-the-shelf channel extender equipment using bus and tag or ESCON interfaces.

B. Remote Mirroring Facility

The data storage system 10 of FIG. 1 is designed to provide the copying of data from a primary data storage system to a physically remote secondary data storage system transparent to the user, and external from any influence of the primary host which is coupled to the primary data storage system. The data storage system 10 is designed to operate in at least two modes, the first being a real-time or synchronous mode wherein the primary and secondary storage systems must guarantee that the data exists and is stored in two physically separate data storage units before input/output completion; that is, before channel end and device end is returned to the primary host. Alternatively, the data storage system 10 is designed to operate in a point-in-time or asynchronous mode wherein the data is copied to the remote or secondary data storage system asynchronously from the time when the primary or local data processing system returns the input/output completion signal (channel end and device end) to the primary host. This eliminates any performance penalty if the communication link between the primary and secondary data storage systems is too slow, but creates the additional needs to manage the situation where data is not identical or in "sync" between the primary and secondary data storage systems.

Thus, in the real time or synchronous mode, the primary data storage system automatically controls the duplication or copying of data to the secondary data storage system controller transparently to the primary host computer. Only after data is safely stored in both the primary and secondary data storage system, as detected by an acknowledgement from the secondary storage system to the primary storage system, does the primary data storage system acknowledge to the primary host computer that the data is synchronized. Should a disaster or facility outage occur at the primary data storage system site, the user will simply need to initialize the application program in the secondary data storage system utilizing a local host (52) or a commercial hotsite CPU or host 56.

The link between the primary and secondary storage system controllers 14 and 46 may be maintained in a unidirectional mode wherein the primary data storage system controller monitors and controls data copying or mirroring. Alternatively, a bi-directional implementation may be used wherein either controller can duplicate data to the other controller, transparently to the host computer. Should a disaster or facilities outage occur, recovery can be automatic with no human intervention since the operational host computer already has an active path (40, 58, 60) to the data through its local controller. While offering uninterrupted recovery, performance will be slower than in an unidirectional implementation due to the over head required to manage intercontroller tasks.

In the second, point-in-time or asynchronous mode of operation, the primary data storage system transparently duplicates data to the secondary data storage system after the primary data storage system acknowledges to the host computer, via channel end and device end, that the data has been written to the storage device and the input/output operation has been completed. This eliminates the performance impact of data mirroring over long distances. Since primary and secondary data are not synchronized, however, the primary data storage system must maintain a log file of pending data which has yet to be written to the secondary data storage device. Such data may be kept on removable, non-volatile media, in the cache memory of the primary or secondary data storage system controller as will be explained below, or in the service processor 34, 62 of the primary or secondary data storage system.

Accordingly, a feature of the data storage system 10 is the ability of a data storage system to control the transfer or copying of data from a primary data storage system to the secondary data storage system, independent of and without intervention from one or more host computers. Most importantly, in order to achieve optimum data mirroring performance, such data mirroring or copying should be performed asynchronously with input/output requests from a host computer. Accordingly, since data will not be immediately synchronized between the primary and secondary data storage systems, data integrity must be maintained by maintaining an index or list of various criteria including a list of data which has not been mirrored or copied, data storage locations for which a reformat operation is pending, a list of invalid data storage device locations or tracks, whether a given device is ready, or whether a device is write-disabled. Information must also be included as to the time of the last operation so that the data may later be synchronized should an error be detected.

A feature of the system of FIG. 1 is that both the primary or secondary data storage systems maintain a table of the validity of data in the other storage system. As disclosed in U.S. Pat. No. 5,206,939 entitled SYSTEM AND METHOD FOR DISK MAPPING AND DATA RETRIEVAL, which is fully incorporated herein by reference, the present system maintains a list or index, utilizing one or more flag bits, in a hierarchical structure, on each physical and logical data storage device.

In the system of FIG. 1, however, such information is kept on both devices for each individual system as well as the other data storage system. Thus, as illustrated in FIG. 2 in the partial list or table 100, each data storage system maintains an indication of write or copy pending 102 of both the primary data (M1) 104, and the secondary data (M2) 106. Similarly, an index is maintained of a pending format change since a disk format change may be accomplished. The format pending bits 108
including a first primary bit 110 and a second secondary bit 112 indicate that a format change has been requested and such change must be made on the disk.

Thus, when a host computer writes data to a primary data storage system, it sets both the primary and secondary bits 104, 106 of the write pending bits 102 when data is written to cache. For these examples, the M1 bit will refer to the primary data storage system and the M2 bit will refer to the secondary data storage system. When the primary data storage system controller's disk adapter writes the data to the primary data storage device, it will reset bit 104 of the write pending indicator bits 102. Similarly, once the secondary data storage system has written the data to the secondary data storage device, the secondary data storage write pending indicator bit 106 will be reset.

The service processors in one embodiment of the present invention will periodically scan the index table for write pending indicator bits and invoke a copy task which copies the data from the primary data storage device to the secondary data storage device. In addition, one or more of the spare index or table bits 114, 116 may be utilized to store other data such as time stamp, etc.

In addition to the write pending and format pending bits described above, the data storage system 10 also includes several additional general purpose flags to assist in error recovery. As shown in FIG. 3, invalid track flags 120 including primary bit 122 and secondary bit 124 are utilized and maintained on each data storage device to indicate that the data storage location such as a track, does not contain valid data. Another background task running on he data storage system such as in the service processor or storage system controller constantly checks invalid track bits on each data storage device, and if a bit is found to be set, the copy task is invoked to copy the data from the known good device to the device with the invalid flag track set. Additional flags may be provided such as the device ready flags 126 including bits 128 and 130 which serve to indicate that the device is ready. Similarly, write disable flags 132 may be provided which indicate that a particular primary device or drive 134 or secondary device or drive 136 can presently not be written to. Data can still be copied to the good or enabled drive and then later copied to the disabled drive. If one drive or device is bad, the present invention will set all tracks of that drive as not valid to later cause a copy of all the data.

Accordingly, each data storage device keeps data validity information about its mirrored device. If for some reason a device is not accessible, either the primary or the secondary device, every new write command goes to the accessible mirrored device along with information that the not accessible device has a track which is not valid. As soon as the non-accessible device becomes accessible, then automatically, as a background operation, the drives re-synchronize. In the case when a specific track is not shown on both the primary and secondary storage system, an indication of such will be assigned and the user will be alerted. A utility operating on the service processors will give the user a report of all the non-valid (out of sync) tracks. This report can be transferred from one site to another over the link 63, FIG. 1, that connects the two service processors 34, 62.

C. Communication Link Options

As introduced above with respect to FIG. 1, the disk adapters 36 and 42 are configured for interconnecting the primary data storage system 14 to the secondary storage system via the high-speed link 40. Further details of various link options are shown in FIGS. 4 to 6.

FIG. 4 shoes a data processing system 210 having a host central processing unit 212, a primary data storage system 214, and a secondary data storage system 246. In the preferred construction shown in FIG. 4, the primary and secondary data storage systems 214, 246 are integrated cached disk arrays having dual, redundant internal and external data links. In particular, the primary date storage system 214 has dual internal busses 238, 239 from a dual-port cache 228, dual channel adapters
226, 227, dual disk adapters 230, 231, and dual link adapters 236, 237. The host 212 at site A is connected to each of the dual channel adapters 226, 227 via respective channel links 218, 219. The secondary data storage system 246 is connected to the dual link adapters 236, 237 in the primary data storage system 214 via respective communicative links 240, 241. The secondary data storage system 246 is also connected to the primary data storage system via dual signal paths 263, 265 from a dual-port service processor 234.

Data storage 220 in the primary data storage system 214 is provided by an array of dual-port disk drives 223a, 223b, 223c, 223d. Each of the disk drives 223a, 223b, 223c, 223d, is connected to each of the disk adapters 230, 231 by a respective fiber channel loop 225, 229. For increased data storage capacity, additional disk drives could be inserted into the fiber channel loops 225, 229, and additional disk adapters could be included in the primary data storage system to accommodate additional fiber channel loops of additional disk drives.

As shown in FIG. 4, the secondary data storage system 246 preferably has the same construction as the primary data storage system 214, and could be linked to the host central processing unit 212 via redundant signal paths 258, 259. The data processing system 210 in FIG. 4 can be configured for remote mirroring from a user interface of the service processor 234 in the primary data storage system. The host central processing unit 212 can also be provided with optional host remote mirroring (RM) software 213 so that the data processing system can be configured and monitored from a user interface of the host central processing unit. Host application programs can also interface with the remote mirroring facility of the data storage systems
214, 246 via the optional host remote mirroring (RM) software 213. An optional host central processing unit 252 could be located at the remote site of the secondary data storage system 246, and linked to each of the primary and secondary data storage systems 214, 246 via redundant signal paths.

The communication links 240, 241 from the dual link adapters 236, 237 are preferably IBM ESCON standard fiber-optic links. An ESCON fiber-optic link, with continuous optical fiber, can link primary and secondary data storage systems spaced by up to 3 kilometers apart. ESCON links between primary and secondary storage units can be extended by repeaters or interfaces to T3 or E3 circuits. In practice, it is desirable to standardize link configurations to two options; namely, a relatively short distance option for distances up to about 60 kilometers (37.5 miles) between the primary and secondary storage units, and a relatively long distance option for distances greater than about 60 kilometers between the primary and secondary data storage systems. In each case, each link adapter has a standard two-port IBM specification LED multimode ESCON interface. It is desirable to provide a minimum of two and a maximum of at least eight link adapters in each data storage system.

Shown in FIG. 5 is the short distance option for interconnecting an integrated cached disk array 301 having link adapters 302, 303 to a remote integrated cached disk array 304 having link adapters 305, 306. Repeaters 307, 308 interface the ESCON channels from each of the link adapters 302, 305 to a private fiber or leased common carrier circuit 309 providing a static connection. In a similar fashion, repeaters 310, 311 interface the ESCON channels from each of the link adapters 303, 306 to a private fiber or leased common carrier circuit 312 providing a static connection. The repeaters 306, 307, 310, 311 are IBM 9032/9033 ESCON Directors or 9036 Remote Channel Extenders. These standard ESCON Directors or Remote Channel Extenders may be used in multiple 20 kilometer hops. In general, for the short distance option, the links can be any combination of multimode fiber, ESCON Directors, Remote Channel Extenders, and single-mode fiber to achieve the maximum link distance of 60 km.

Shown in FIG. 6 is the long distance option for interconnecting an integrated cached disk array 321 having link adapters 322, 323 to an integrated cached disk array 324 having link adapters 325, 326. ESCON to T3/E3 converters 327, 328 interface the ESCON channels from each of the link adapters 322, 325 to a T3 or E3 circuit 329. In a similar fashion, repeaters 330, 331 interface the ESCON channels from each of the link adapters 303, 306 to a T3 or E3 circuit 332. A suitable ESCON to T3/E3
converter may include Data Switch Corporation Model 9800 MAX (Multiple Architecture Extender). The 9000 MAX accepts up to four ESCON inputs, and multiplexes the data across 1 or 2 lines. T3 and E3 are copper or fiber-based telecommunications circuit. T3 is available in North America, and E3 is available in Europe. T3 has a bandwidth of 44.5 megabits per second, and E3 has a bandwidth of 34.5 megabits per second. A T3 or E3 circuit is sometimes referred to as "broad band". A T3/E3 circuit can be "fragmented", subdivided for multiple application or user access, or be dedicated point-to-point.

Data channels between a host and a storage system remote from the host can be constructed in a fashion similar to the links shown in FIGS. 5 or 6.

D. Initial Synchronization

Once the physical links are established between the primary and secondary data storage systems, and the user specifies which logical storage devices or volumes are to be remotely mirrored, appropriate microcode is loaded into the data storage systems. It is also possible that the primary and secondary logical volumes could also be configured for local mirroring for enhanced redundancy. Alternatively, local redundancy could employ techniques for distributing the data bits of each byte or word of data in a logical device or volume across a multiplicity of physical disk drives in various ways known as levels of RAID (redundant arrays of inexpensive disks).

RAID techniques are described in the following publications: Patterson et al., "A Case for Redundant Arrays of Inexpensive Disks (RAID)," Report No. UCB/CSD 87/391, Computer Science Division (EECS), University of California, Berkeley, Calif., December 1987 (pages 1 to 24); Patterson et al., "Introduction to Redundant Arrays of Inexpensive Disks (RAID)," COMPCON 89 Proceedings, Feb. 27-Mar. 3, 1989, IEEE Computer Society, pp. 112-117; Ousterhout et al., "Beating the I/O Bottleneck: A Case for Log-Structured File Systems," Operating Systems Review, Vol. 23, No. 1, ACM Press, January, 1989, pp. 11-28; Douglis et al., "Log Structured File Systems," COMPCON 89 Proceedings, Feb. 27-Mar. 3, 1989, IEEE Computer Society, pp. 124-129; and Rosemblum et al., "The Design and Implementation of a Log-Structured File System," ACM Transactions on Computer Systems, Vol. 1, February 1992, pp. 26-52; which are all incorporated herein by reference.

As soon at the communication links are established to interconnect the primary and secondary data storage systems, synchronization of the primary and secondary storage devices or logical volumes begins, and data is copied from the primary (R1) devices to the secondary (R2) devices. While this initial synchronization is occurring, host application input/output may be addressed to the primary (R1) devices. Typically, this application input/output is given precedence over the initial synchronization activity.

E. Multiple Simultaneous Operating Modes for the Remote Mirroring Facility

It is advantageous to provide the remote mirroring facility in the system 210 of FIG. 4 with multiple simultaneous operating modes best suited for the purposes of the desired remote mirroring. For example, remote mirroring may be used for data migration as well as for disaster recovery, and specific operating modes will be described that are best suited for data migration, and others will be described that are best suited for disaster recovery. Data migration, for example, typically occurs when a data center is moved from one geographic location to another, or when an old data storage system is replaced with a new data storage system.

Specific operating modes will also be described that are best suited for particular application programs. Different application programs, for example, may have different requirements for critically of data integrity. Certain application programs may have specific procedures, such as transaction processing or journaling facilities, for ensuring data integrity relatively independent of the data integrity of the data storage systems.

The suitability of remote mirroring may also depend on the particular use or purpose of a dataset. Data bases, logs, catalogs, system residence volumes, and program libraries are excellent candidates for remote mirroring. Multiple logs when placed on separate logical volumes on different physical devices also aid business operations recovery in the event of a disaster. Page, spool, work, and sort datasets, however, are poor remote mirroring candidates as they are write-intensive often to only a small number of volumes.

To provide multiple simultaneous remote mirroring operating modes for specific applications, the remote mirroring facility defines an operating mode for each logical volume of data in the storage devices in the primary and secondary data storage systems 214, 246. Each logical volume may include a number of logical tracks of data and may reside on one or more disk drives in either the primary or secondary data storage system 214, 246.

Each logical volume has a logical volume type that is either primary, secondary, or local. A local logical volume does not participate in remote mirroring. A pair (R1, R2) of respective primary (R1) and secondary (R2) logical volumes participates in remote mirroring according to either a synchronous mode, a semi-synchronous mode, an adaptive copy--write pending mode, or an adaptive copy--risk mode, as will be further described below.

The operational modes are selectable at the logical volume level based on the performance, distance, and speed of recovery requirements. All primary (R1) volumes are configured for either the synchronous or semi-synchronous mode. These two modes are considered to be pre-determined remote mirroring modes. In addition, the primary (R1) volumes (all, individual, or a range) may also be configured for the adaptive copy--write pending or adaptive copy--disk mode. Each volume configured for adaptive copy also has an associated "skew" parameter. In the adaptive copy--write pending mode, this skew parameter is the maximum write pending threshold. In the adaptive copy--disk mode, this skew parameter is the maximum invalid tracks threshold. This skew value may he set to the same value for all adaptive copy volumes or be a different value for each adaptive copy volume. The adaptive copy mode and its skew value may be enabled (or disabled) for individual remotely mirrored pairs or all remotely mirrored pairs using remote mirroring commands.

(1) Synchronous Mode

In the synchronous mode, data on the primary (R1) and secondary (R2) volumes are always fully synchronized at the completion of an I/O sequence. The data storage system containing the primary (R1) volume informs the host that an I/O sequence has successfully completed only after the data storage system containing the secondary (R2) volume acknowledges that it has received and checked the data.

In particular, when the data storage system containing the primary (R1) volume has valid data in cache destined for a secondary (R2) volume, a link adapter transfers data over its link path to the cache in the data storage system housing the secondary (R2) volume. This data transfer occurs while the data storage system containing the primary (R1) volume continues to process input/output commands. If the data storage system containing the primary (R1) volume does not receive acknowledgment of a successful transfer from the other data storage system within a timeout period or another failure occurs that prevents the data transfer, the data storage system containing the primary (R1) volume sends a "unit check" with appropriate sense bytes to the host.

In a CKD environment, the data storage system containing the primary (R1) volume sends channel end (CE) and device end (DE) to the host after each write to the volume with the exception of the last write in the channel command word (CCW) chain. On the last write, the data storage system sends only CE to the host. When the data storage system containing the secondary (R2) volume acknowledges and checks receipt of the last write in the chain, the data storage system containing the primary (R1) volume sends DE to the host and the host considers the input/output complete and starts the next input/output operation.

In an open systems environment, the data storage system containing the primary (R1) volume handles each input/output command separately and informs the host of successful completion when the data storage system containing the secondary (R2) volume acknowledges and checks receipt of the data. That is, the data storage system containing the primary (R1) volume disconnects from the channel and informs the host of successful completion of the input/output operation only after confirming that the data resides in cache in both data storage systems. If a problem occurs with data synchronization, the data storage system containing the primary (R1) volume sends a "unit check" with appropriate sense bytes to the host. This causes the host to retry the input/output operation. These actions maintain data integrity and ensure that two copies of the data exist real-time in both systems before the input/output completes.

The synchronous mode is recommended primarily for the short distance option of FIG. 5. In normal operation, this mode will have an impact on write performance to primary (R1) volumes. This performance impact is due to overhead associated with remote data transfer, fiber latency, and acknowledgment of the synchronous operation.

(2) Semi-synchronous Mode

In the semi-synchronous mode, the remotely mirrored volumes (R1, R2) are always synchronized between the primary (R1) and the secondary (R2) prior to initiating the next write operation to these volumes. The data storage system containing the primary (R1) volume informs the host of successful completion after each write operation.

When the data storage system containing the primary (R1) volume has valid data in cache destined for a secondary (R2) volume, a link adapter transfers data via an available link path to the cache in the data storage system containing the secondary (R2) volume. This data transfer occurs while the data storage system containing the primary (R1) volume continues to perform additional channel commands. If the host issues a new write operation for a primary (R1) volume with a write pending status, the data storage system containing the primary (R1) volume disconnects from the host channel and returns a "non-immediate retry" message. The data storage system containing the primary (R1) volume then starts another input/output operation on another channel. When the write pending status is cleared (write completed and acknowledged and checked from the secondary (R2) volume), the data storage system containing the primary (R1) volume reconnects to the channel and continues processing the write operation on the channel from which it disconnected.

The semi-synchronous mode is recommended primarily for the long distance option of FIG. 6. The semi-synchronous mode is designed for situations needing high performance at the data storage system containing the primary (R1) volume and tolerating a gap of up to one input/output (worst case) in data synchronization. Although write operations can be held up due to synchronization between primary (R1) and secondary (R2) volumes, read operations continue uninterrupted.

The semi-synchronous mode is most suitable for page, spool, work, and sort datasets. In some cases, spreading these datasets across multiple physical devices may alleviate any performance impact due to a high number of writes.

(3) Channel Adapter Control Logic for the Pre-determined Modes

Turning now to FIGS. 7 and 8, there is shown a flowchart of channel adapter control logic for the synchronous and semi-synchronous modes. In the preferred implementation, this control logic is specified by programming for microprocessors in the channel adapters.

In FIG. 7, a first step 401 is reached when the channel adapter receives a channel command from the host requesting data access to a volume. It is assumed that the host is not requesting direct access to a secondary (R2) volume in the data storage system containing the channel adapter. The host may request direct access to a secondary (R2) volume during recovery operations, which are described below. It is also assumed that the channel command is not in a chain of multiple channel commands. The chaining of multiple channel commands is described below with reference to FIG. 19.

In the first step 401 of FIG. 7, execution branches to step 402 for a read access. In step 402, the channel adapter accesses configuration information, and continues to step 403 if the host is requesting access to a local volume. Preferably, a separate copy of the configuration information is stored in local memory in each of the channel adapters and link adapters. This configuration information identifies whether a volume is local, primary, or secondary, and for each primary or secondary volume, identifies the other volume in the remotely mirrored volume pair.

In step 403, the channel adapter accesses the cache. If the data requested by the host is not in the cache, then the data is fetched by a disk adapter from disk storage in the data storage system, and loaded into the cache. Then, in step 404, the channel adapter transmits the data and a device end signal to the host, and the channel adapter has finished the task of servicing the channel command.

If the host channel command is requesting data in the primary (R1) volume of a remotely mirrored pair, then execution branches from step 402 to step 405. In step 405, execution branches to step 403 unless the data storage system is in the synchronous mode. For modes other than the synchronous mode, the reading of data from a primary (R1) volume is normally similar to the reading of data from a local volume; in either case, the requested data is fetched without delay from the cache or disk in step 403. Under the abnormal condition of the data being entirely absent from the data storage system due to a disk drive failure, however, a request for data access to a primary (R1) volume can be satisfied by obtaining the requested data from the secondary volume (R2) in the remote data storage system. The handling of such an abnormal condition is discussed below in connection with data recovery procedures.

In step 406, when a remote write is not pending to the secondary (R2) of the requested mirrored volume, execution also branches to step 403 to fetch the requested data from the cache or disk. When a remote write is pending to the secondary (R2) of the requested mirrored volume, however, execution continues to step 407 to suspend the current read task until the remote data storage system acknowledges completion of the pending remote write. Preferably, tasks suspended while waiting for completion of a pending remote write are placed on a first-in first-out (FIFO) queue of suspended tasks, and when the remote data storage system acknowledges completion of the pending remote write, any waiting tasks in queue of suspended tasks are serviced in the order in which the tasks were placed in the queue. Once the remote data storage system acknowledges completion of the pending remote write, and no remote write to the secondary (R2) of the mirrored volume is pending, as tested in step
406, execution branches to step 403 to fetch the requested data from the cache or disk.

When the host has requested a write access, execution continues from step 401 to step 408. In step 408, execution branches to step 409 when the host has requested a write access to a volume that is local. In step 409, data from the host is written to cache, and the track tables are updated to reflect that the old data on disk is invalid in view of the new data from the host, and that a write operation to disk is pending for the invalid track or tracks on disk.

Then in step 410, a device end (DE) signal is returned to the host to signal completion of the write operation. The signaling of the completion of a write operation before the data is actually written to disk is a well-known technique called "fast write." Semiconductor random-access memory containing the write data is backed-up by a battery sufficient to power the memory and some disk drives while the write data is transferred to the disk drives in the event of a power failure.

When the host has requested a write operation to a volume defined as a mirrored volume pair, execution continues from step 408 to step 411. In step 411, execution continues to step 412 when a remote write to the secondary (R2) of the remotely mirrored volume is pending. In step 412, the current write task is temporarily suspended, while awaiting receipt from the remote data storage system of acknowledgement of completion of the pending remote write, as tested in step 411. When no remote writes to the secondary (R2) of the remotely mirrored volume are pending, execution branches from step 411 to step 414 in FIG. 8.

In step 414 of FIG. 8, the data from the host is written to the cache, and the track tables are updated to indicate that the track or tracks for the new data in disk for the primary (R1) volume are invalid and have a pending write operation to disk, and that the track or tracks for the new data are invalid in the secondary (R2) of the remotely mirrored volume and have a pending write to the cache in the remote data storage system. Due to the incorporation of the "fast write" technique of acknowledging a write to a secondary (R2) volume when the update is written to cache of the data storage system containing the secondary volume, the remote "invalid" and "write pending" status for the secondary (R2) volume in the track tables of the data storage system containing the corresponding primary (R1) volume refers to the status of the secondary (R2) volume in cache or on disk; in particular, the remote "write pending" status indicates a pending write over the link to the cache in the data storage system containing the secondary (R2) volume. When the "fast write" technique is used, it is still necessary, for carrying out the local destage or write back operation, for each data storage system to record, for each track or data record, an indication of whether a local destage operation is pending, and such a local destage operation is pending when the track or record is valid and is in cache but the disk drives do not have valid data for the track or record.

Next, in step 415, the write data from the host is written to a first-in, first-out (FIFO) link transmission queue (504 in FIG. 18) for transmission by a link adapter to the remote data storage system. Preferably, the entries in the queue contain pointers to the data in cache. When a link adapter becomes available, it services this FIFO queue by transmitting the data identified by the entry at the head of the queue across the link to the remote data storage system.

Next, in step 416, execution branches to step 417 when the data storage system is not in the synchronous mode. In step 417, the channel adapter transmits a device end (DE) signal to the host, and execution continues to step 418. Execution also continues to step 418 from step 416 when the data storage system is in the synchronous mode.

In step 418, the current write task is suspended, until the remote data storage system has received the write data, written the data in its cache, and has acknowledged completion of the remote write operation. In the short distance option, the remote acknowledgement should be received just before a next remote write task sends data over the link, and therefore it may be feasible for the link adapter to poll for the remote acknowledgement. In the long distance option, the next remote write task may end data over the link well before the acknowledgement is received, so that receipt of the acknowledgement causes an interrupt re-activating the suspended write task. Once the data storage system receives the acknowledgement of completion of the remote write, as tested in step 419, execution continues to step 420. In step 420, the track tables are updated to indicate completion of the remote write to the cache of the secondary (R2) volume in the remotely mirrored volume pair, so that the track or tracks of the new write data are valid in the secondary (R2) volume.

From the control flow in FIGS. 7 and 8, it is clear that when a host writes data to a remotely mirrored volume, the following sequence of events takes place in the synchronous mode: data is written to the cache of the data storage system containing the primary (R1) volume (step 414); an entry is placed in the FIFO lin