Cross tagging to data for consistent recovery

1. A system for consistent data recovery, comprising a storage area network (SAN) of initiator nodes and target nodes, wherein initiator nodes issue I/O requests and target nodes respond to I/O requests, the SAN being represented logically as a plurality of consistency groups of nodes, G 1 , . . . , Gn, each consistency group representing:

(A) at least one logical storage unit (LU), each LU acting as a target node within the SAN;

(B) at least one host device, each host device acting as an initiator within the SAN, each host device comprising (B.1) a transmitter for issuing I/O requests to at least one LU, and each host device being associated with (C) a host device agent, the host device agent comprising:

(C.1) an interceptor for intercepting I/O requests issued by the host device and performing one or more of sending the intercepted I/O requests to the at least one LU, redirecting the intercepted I/O request to another LU, splitting the intercepted I/O requests to enable sending of the intercepted I/O requests first to a data protection agent prior to transmission to the at least one LU; and

(C.2) a router for forwarding the intercepted I/O requests to the data protection appliance;

(D) a data protection appliance (DPA) acting as both an initiator node and a target node within the SAN, the DPA comprising:

(D.1) a receiver for receiving a current I/O request forwarded by an agent router for a host device;

(D.2) a router for forwarding the current I/O request to a replication facility;

(D.3) an ACK transmitter for sending an acknowledgement for an I/O request to an agent, wherein said agent router is operable to halt forwarding further host device I/O requests until the acknowledgement for a current I/O request is received from the DPA, and to resume forwarding after the acknowledgement for the current I/O request is received; and

(D.4) a tag initiator for issuing an instruction for the replication facility to tag a consistency group at a current point in time; and

(E) a replication facility comprising:

(E.1) a receiver for receiving I/O requests forwarded by said DPA router; and

(E.2) a data recoverer for restoring the LUs to a tagged state, based on I/O requests received by said DPA receiver,

the system further comprising (F) a DPA manager operable to send requests to said DPAs of G 1 , . . . , Gn,

(r 1 ) to stop their ACK transmitters sending acknowledgements for I/O requests to said host device agents of G 1 , . . . , Gn, and

(r 2 ) to tag G 1 , . . . , Gn.

2. The system of claim 1 wherein said DPAs of G 1 , . . . , Gn are operable to send back confirmations to said DPA manager upon completion of said DPA manager's requests (r 1 ) and (r 2 ), and wherein said DPA manager, after receiving confirmations from said DPAs of G 1 , . . . , Gn, is operable to send a request to said DPAs of G 1 , . . . , Gn,

(r 3 ) to restart their ACK transmitters sending acknowledgements of I/O requests to said host device agents of G 1 , . . . , Gn.

3. The system of claim 2 wherein said DPA manager further comprises a timer for setting a time limit, and wherein said DPA manager is operable to report a cross-tagging success if said DPA manager receives confirmations from said DPAs of G 1 , . . . , Gn, within the time limit set by said DPA manager timer.

4. The system of claim 3 wherein each DPA further comprises a timer for setting a time limit, and wherein a DPA is operable to start its ACK transmitter sending acknowledgements of I/O requests to host device agents after the time limit set by said DPA timer has elapsed.

5. The system of claim 2 wherein each DPA of G 1 , . . . , Gn further comprises a timer for setting a time limit, and wherein each DPA of G 1 , . . . , Gn is operable to start sending acknowledgements of I/O requests to host devices after the time limit set by said DPA timer has elapsed, and wherein each DPA of G 1 , . . . , Gn is operable to send a notification to said DPA manager as to whether or not its ACK transmitter had already restarted sending acknowledgements when the DPA received the DPA manager's request (r 3 ) to restart them.

6. The system of claim 1 wherein said DPA manager is operable to report a cross-tagging success if said DPA manager receiver receives notifications from said DPAs of G 1 , . . . , Gn that their ACK transmitters had not already restarted sending acknowledgements when they received the DPA manager's request (r 3 ) to restart them.

7. The system of claim 1 wherein each of said host device agents of G 1 , . . . , Gn comprises a timer for setting a time limit, and wherein a host device agent router is operable to resume forwarding I/O requests after the time limit set by said host device agent timer has elapsed.

8. The system of claim 1 wherein a plurality of said DPAs are configured as a computer cluster.

9. A method for consistent data recovery, comprising issuing requests from a data protection appliance (DPA) manager to at least one DPA within at least one consistency group, G 1 , . . . , Gn,

(r 1 ) to stop sending acknowledgements of I/O requests to host device agents of G 1 , . . . , Gn, and

(r 2 ) to tag G 1 , . . . , Gn, wherein a tag is used to mark a consistency group at a specific point in time,

wherein the consistency groups G 1 , . . . , Gn represent logical parts of a storage area network, and each consistency group G 1 , . . . , Gn represents a DPA, at least one host device, at least one LU and at least one replication facility, the replication facility being operable to restore the at least one LU to its former state at the time of the tag, and wherein each host device is associated with a host device controller that is operable to:

forward the host device I/O requests;

halt such forwarding when it does not receive an acknowledgement of a current I/O request from a DPA; and

resume such forwarding after it does receive an acknowledgement of the current I/O request from the DPA.

10. The method of claim 9 further comprising performing, by each DPA of G 1 , . . . , Gn, the DPA manager's requests (r 1 ) and (r 2 ), said performing comprising:

stop sending acknowledgements of I/O requests to the host device controllers of its consistency group; and

tagging its consistency group.

11. The method of claim 10 further comprising:

sending, by each DPA of G 1 , . . . , Gn, a confirmation to the DPA manager after said performing the DPA manager's requests; and

issuing, by the DPA manager after the DPA manager receives confirmations from all DPAs of G 1 , . . . , Gn, requests to the DPAs of G 1 , . . . , Gn,

(r 3 ) to restart sending acknowledgements of I/O requests to the host device controllers of G 1 , . . . , Gn.

12. The method of claim 11 further comprising:

setting a time limit, by the DPA manager; and

reporting, by the DPA manager, a cross-tagging success if the DPA manager receives confirmations from all DPAs of G 1 , . . . , Gn within the time limit set by the DPA manager.

13. The method of claim 12 further comprising setting, by each DPA, a time limit, and wherein each DPA restarts sending acknowledgments of I/O requests to host device controllers after the time limit set by the DPA has elapsed.

14. The method of claim 13 further comprising sending a notification, by each DPA of G 1 , . . . , Gn, to the DPA manager as to whether or not it had already restarted sending acknowledgements of I/O requests when it received the DPA manager's request (r 3 ) to restart the sending.

15. The method of claim 14 further comprising reporting, by the DPA manager, a cross-tagging success if notifications are received from the DPAs of G 1 , . . . , Gn that they had not already restarted sending acknowledgements of I/O requests when they received the DPA manager's request (r 3 ) to restart the sending.

16. The method of claim 10 further comprising restoring, by each replication facility, the LUs of its consistency group to their states at the time of said tagging by the DPA of its consistency group.

17. The method of claim 9 further comprising setting, by each host device controller, a time limit, wherein the host device controller resumes forwarding I/O requests after the time limit set by the host device agent has elapsed.

18. A computer-readable storage medium storing program code for causing a data protection appliance (DPA) manager to issue requests to at least one DPA within at least one other consistency group, G 1 , . . . , Gn,

(r 1 ) to stop sending acknowledgements of I/O requests to host device agents of G 1 , . . . , Gn, and

(r 2 ) to tag G 1 , . . . , Gn, wherein a tag is used to mark a consistency group at a specific point in time,

wherein the consistency groups G 1 , . . . , Gn represent logical parts of a storage area network, and each consistency group G 1 , . . . , Gn represents a DPA, at least one host device, at least one LU and at least one replication facility, the replication facility being operable to restore the at least one LU to its former state at the time of the tag, and wherein each host device is associated with a host device agent that is operable to:

forward the host device I/O requests;

halt such forwarding when it does not receive an acknowledgement of a current I/O request from a DPA; and

resume such forwarding after it does receive an acknowledgement of the current I/O request from the DPA.

19. A system for data protection, comprising a network represented logically as a plurality of consistency groups, G 1 , . . . , Gn, each consistency group representing:

at least one host device for issuing I/O requests, and each host device being associated with a host device controller for processing I/O requests issued by the host device; and

a data protection appliance (DPA), communicating with said at least one host device, comprising:

a quiesce controller operable to cause said at least one host device controller to halt processing I/O requests and to cause said at least one host device controller to resume processing I/O requests; and

a tag initiator for marking a specific consistency group at a specific point in time;

the system further comprising a DPA manager operable to send requests to said DPAs of G 1 , . . . , Gn,

(r 1 ) via their quiesce controllers, to cause said host device controllers of G 1 , . . . , Gn to halt processing I/O requests, and

(r 2 ) to initiate tags in G 1 , . . . , Gn.

20. The system of claim 19 wherein said DPAs of G 1 , . . . , Gn are operable to send back confirmations to said DPA manager upon completion of said DPA manager's requests (r 1 ) and (r 2 ).

21. The system of claim 20 wherein said DPA manager, after receiving confirmations from said DPAs of G 1 , . . . , Gn is operable to send a request to said DPAs from G 1 , . . . , Gn,

(r 3 ) via their quiesce controllers, to cause said host device controllers of G 1 , . . . , Gn to resume processing I/O requests.

22. The system of claim 20 wherein said DPA manager further comprises a timer for setting a time limit, and wherein said DPA manager is operable to report a cross-tagging success if said DPA manager receives confirmations from said DPAs of G 1 , . . . , Gn, within the time limit set by said DPA manager timer.

23. The system of claim 22 wherein each DPA further comprises a timer for setting a time limit, and wherein said DPA quiesce controller is operable to cause a host device controller to resume processing I/O requests after the time limit set by said DPA timer has elapsed.

24. The system of claim 20 wherein each DPA of G 1 , . . . , Gn further comprises a timer for setting a time limit, and wherein each DPA quiesce controller is operable to cause its host device controllers to resume processing I/O requests after the time limit set by said DPA timer has elapsed.

25. The system of claim 24 wherein each DPA is further operable to send a notification to said DPA manager as to whether or not its quiesce controller had already caused its host controllers to resume processing I/O requests when said DPA receiver received said DPA manager's request (r 3 ) to cause to resume.

26. The system of claim 25 wherein said DPA manager is operable to report a cross-tagging success if said DPA manager receives notifications from said DPAs of G 1 , . . . , Gn that their quiesce controllers had not already caused their host controllers to resume processing I/O requests when they received said DPA manager's request (r 3 ).

27. The system of claim 19 wherein each of said host device controllers of G 1 , . . . , Gn comprises a timer for setting a time limit, and wherein a host device controller is operable to resume processing I/O requests after the time limit set by said host device controller timer has elapsed.

28. The system of claim 19 wherein a plurality of said DPAs are configured as a computer cluster.

29. The system of claim 19 wherein said host device controller halts processing I/O requests if said quiesce controller stops sending acknowledgements to said host device controller, and wherein said host device controller resumes processing I/O requests if said quiesce controller restarts sending acknowledgements to said host device controller.

30. The system of claim 19 wherein said host device controller halts processing I/O requests if said quiesce controller sends a halt instruction to said host device controller, and wherein said host device controller resumes processing I/O requests if said quiesce controller sends a resume instruction to said host device controller.

31. The system of claim 19 wherein the DPA manager is one of the DPAs of G 1 , . . . , Gn.

32. The system of claim 19 wherein the DPA manager is a separate appliance from G 1 , . . . , Gn.

33. A method for data protection, comprising issuing requests from a data protection appliance (DPA) manager, to at least one DPA within at least one consistency group, G 1 , . . . , Gn, the request comprising:

(r 1 ) to cause host device controllers of G 1 , . . . , Gn to halt processing I/O requests, and

(r 2 ) to initiate tags in G 1 , . . . , Gn, wherein a tag is used to mark a consistency group at a specific point in time,

wherein the consistency groups G 1 , . . . , Gn represent logical parts of a network, and each consistency group G 1 , . . . , Gn represents a DPA and at least one host device, and wherein each host device is associated with a host device controller that processes the host device I/O requests, and that can halt such processing and that can resume such processing.

34. The method of claim 33 further comprising performing, by each DPA of G 1 , . . . , Gn, the DPA manager's requests (r 1 ) and (r 2 ), said performing comprising:

causing the host device controllers of its consistency group to halt processing I/O requests; and

initiating a tag in its consistency group.

35. The method of claim 34 wherein said causing the host device controllers to halt processing I/O requests comprises stopping to send acknowledgements to the host device controllers.

36. The method of claim 34 wherein said causing the host device controllers to halt processing I/O requests comprises sending halt commands to the host device controllers.

37. The method of claim 34 further comprising sending, by each DPA of G 1 , . . . , Gn, a confirmation to the DPA manager after said performing the DPA manager's requests.

38. The method of claim 37 further comprising issuing, by the DPA manager after the DPA manager receives confirmations from all DPAs of G 1 , . . . , Gn, requests to the DPAs of G 1 , . . . , Gn,

(r 3 ) to cause the host device controllers within G 1 , . . . , Gn to resume processing I/O requests.

39. The method of claim 37 further comprising:

setting a time limit, by the DPA manager; and

reporting, by the DPA manager, a cross-tagging success if the DPA manager receives confirmations from all DPAs of G 1 , . . . , Gn within the time limit set by the DPA manager.

40. The method of claim 39 further comprising setting a time limit, by each DPA, and wherein the DPA causes a host device controller to resume processing I/O requests after the time limit set by the DPA has elapsed.

41. The method of claim 38 further comprising setting, by each DPA, a time limit, wherein the DPA causes a host device controller to resume processing I/O requests after the time limit set by the DPA has elapsed.

42. The method of claim 41 further comprising sending a notification, by each DPA of G 1 , . . . , Gn, to the DPA manager as to whether or not it had already caused its host device controllers to resume processing I/O requests when it received the DPA manager's request (r 3 ) to cause them to resume.

43. The method of claim 42 further comprising reporting, by the DPA manager, a cross-tagging success if notifications are received from the DPAs of G 2 , . . . , Gn that they had not already caused their host device controllers to resume processing I/O requests when they received the DPA manager's request (r 3 ) to cause them to resume.

44. The method of claim 33 further comprising setting, by each host device controller, a time limit, wherein the host device controller resumes processing I/O requests after the time limit set by the host device controller has elapsed.

45. A computer-readable storage medium storing program code for causing a data protection appliance (DPA) manager to issue requests to at least one DPA within at least one consistency group, G 1 , . . . , Gn, the requests comprising:

(r 1 ) to cause host device controllers of G 1 , . . . , Gn to halt processing I/O requests, and

(r 2 ) to initiate tags in G 1 , . . . , Gn, wherein a tag is used to mark a consistency group at a specific point in time,

wherein the consistency groups G 1 , . . . , Gn represent logical parts of a network, and each consistency group G 1 , . . . , Gn represents a DPA and at least one host device, and wherein each host device is associated with a host device controller that processes the host device I/O requests, and that can halt such processing and that can resume such processing.

46. A system for data recovery to a crash consistent state of data of a plurality of consistency groups stored in a storage sub-system, comprising:

(A) a storage area network (SAN) of initiator nodes and target nodes, wherein initiator nodes issue I/O requests and target nodes respond to I/O requests, the SAN comprising:

(A.1) at least one storage subsystem acting as a SAN target node, comprising:

(A.1.1) at least one addressable physical storage unit; and

(A.1.2) at least one controller operable to provide access to data stored in said at least one addressable physical storage unit, to at least one host acting as a SAN initiator node through at least one logical storage unit (LU), each LU having an identifier; and

(A.2) at least one host acting as a SAN initiator node, operable to issue I/O requests to said at least one LU through at least one host device, each host device having an identifier;

(B) at least one data processing unit, each data processing unit comprising:

(B.1) a receiver for receiving I/O requests issued by at least one host to at least one LU;

(B.2) a router for forwarding a current I/O request to a data protection facility; and

(B.3) a transmitter for sending an acknowledgement for an I/O request to a host device of said at least one host, wherein each of said at least one host is operable to halt processing further I/O requests through a host device until the acknowledgement is received from the data processing unit for a current I/O request, and to resume processing I/O requests through a host device after the acknowledgement is received;

(C) at least one consistency group, each consistency group being a logical entity comprising:

(C.1) an identifier;

(C.2) at least one of the LU identifiers; and

(C.3) at least one of the host device identifiers;

(D) a tag initiator for issuing an instruction for the data protection facility to store a consistency group identifier and a tag identifier; and

(E) a data protection facility comprising:

(E.1) a receiver for receiving I/O requests forwarded by said at least one router; and

(E.2) a journal processor for storing write I/O requests issued to at least one LU, each write I/O request comprising an identifier of the LU to which the I/O request was issued and an identifier of the sequential order of the write request; and

(E.3) a data recoverer for processing the data stored by said journal processor and the data stored in said at least one LU, or a copy thereof, so to enable at least one SAN initiator node to process the at least one LU, or a copy thereof, in a state it was in at least one point in time in the past,

wherein one of the data processing units of the system further comprises (B.4) a data protection manager operable to send requests, via said transmitter, to at least one data processing unit,

(r 1 ) to halt host devices of at least one consistency group handled by that data processing unit, and

(r 2 ) to tag a consistency group.