Techniques for mirroring data within a shared virtual memory system

• US 7,774,645 B1
• Filed: 03/29/2006
• Issued: 08/10/2010
• Est. Priority Date: 03/29/2006
• Status: Active Grant

• Current Assignee: Emc IP Holding Company LLC (Dell Technologies Inc.)
• Original Assignee: EMC Corporation (Dell Technologies Inc.)
• Inventors: DesRoches, David W., Clark, Roy
• Primary Examiner(s): Baderman; Scott T
• Assistant Examiner(s): SCHELL, JOSEPH O
• Application Number:
  US11/393,438
• Time in Patent Office: 1,595 Days
• Field of Search: 714/6, 714/12, 714/15, 714/7, 714/8, 714/11, 714/13, 711/147, 711/148, 711/161, 711/162
• US Class Current: 714/12
• CPC Class Codes: G06F 11/1666 where the redundant component is memory or memory area
• CPC Class Codes: G06F 11/2058 using more than 2 mirrored copies
• CPC Class Codes: G06F 11/2076 Synchronous techniques
• CPC Class Codes: G06F 12/10 Address translation

Abstract

A technique protects shared data in a local device having local memory. The technique involves observing a page table entry (PTE) on the local device. The PTE is stored in a page table used for managing virtual to physical address translations, tracking page modifications and handling page faults between semiconductor memory and magnetic disk drive memory on the local device. The technique further involves leaving a mirroring routine inactive on the local device when the PTE indicates that shared data corresponding to the PTE has not been modified on the local device. The mirroring routine is configured to copy the shared data from the local memory to a remote device. The technique further involves activating the mirroring routine to mirror the shared data from the local memory to the remote device when the PTE indicates that shared data corresponding to the PTE has been modified on the local device.

Claims

1. In a local computerized device having local physical memory, a method for protecting shared data, the method comprising:

observing a page table entry on the local computerized device, the page table entry being stored in a page table used for managing virtual to physical address translations, tracking page modifications and handling page faults between semiconductor memory and magnetic disk drive memory on the local computerized device;

when the page table entry indicates that shared data corresponding to the page table entry has not been modified on the local computerized device, leaving a data mirror routine inactive on the local computerized device, the data mirror routine being configured to copy the shared data from the local physical memory to a remote computerized device; and

when the page table entry indicates that shared data corresponding to the page table entry has been modified on the local computerized device, activating the data mirror routine to mirror the shared data from the local physical memory to the remote computerized device;

wherein the local physical memory of the local computerized device includes a local non-mirror memory area and a local mirror memory area;

wherein remote physical memory of the local computerized device includes a remote non-mirror memory area and a remote mirror memory area;

wherein the local non-mirror memory area and the remote non-mirror memory area store different data to independently run, on the local and the remote computer devices, different applications performing different tasks over different fault domains in an ongoing manner;

wherein the local mirror memory area and the remote mirror memory area maintain mirroring copies of the shared data to provide fault tolerant redundancy for the shared data across the different fault domains while the local and the remote computer devices independently run the different applications performing the different tasks over the different fault domains in the ongoing manner; and

wherein the data mirror routine, in response to activation when the page table entry indicates that the shared data corresponding to the page table entry has been modified on the local computerized device, is constructed and arranged to (i) impose a lock on the shared data, (ii) perform a copy operation which copies the shared data from the local physical memory to the remote computerized device while concurrently maintaining the lock on the shared data, and (iii) release the lock after the copy operation is completed.

2. A method as in claim 1 wherein a computerized system is formed by multiple computerized devices which include the local computerized device and the remote computerized device;

and wherein the method further comprises;

pre-defining, for each computerized device of the multiple computerized devices forming the computerized system, another computerized device of the multiple computerized devices to operate as a mirror for shared data on that computerized device, the remote computerized device being configured to operate as a mirror for the shared data on the local computerized device.

3. A method as in claim 2, further comprising:

suffering a failure which (i) prevents the local computerized device from providing access to the shared data on the local physical memory and (ii) enables access to the shared data on the remote computerized device, the shared data on the remote computerized device having been copied to the remote computerized device by the data mirror routine.

4. A method as in claim 2 wherein observing the page table entry includes:

inspecting a modified bit to determine whether the shared data corresponding to the page table entry has been modified on the local computerized device, the modified bit indicating that the shared data (i) has not been modified when the modified bit is unasserted and (ii) has been modified when the modified bit is asserted.

5. A method as in claim 4 wherein activating the data mirror routine to mirror the shared data from the local physical memory to the remote computerized device includes:

in response to a determination that the modified bit has been asserted, mirroring the shared data from the local physical memory to the remote computerized device.

6. A method as in claim 5 wherein the computerized system is formed by N computerized devices;

wherein a page of the local physical memory is divided into Q extents;

wherein the shared data resides in a particular extent of the Q extents; and

wherein mirroring the shared data from the local physical memory to the remote computerized device includes;

copying the shared data from the particular extent of the Q extents to a corresponding extent in physical memory of the remote computerized device.

7. A method as in claim 1 wherein the local mirror memory area includes (i) a first page storing first data which is updated by a first application running on the local computerized device and (ii) a second page storing second data which is updated by a second application running on the remote computerized device, the first and second applications running concurrently to perform at least some of the different tasks over the different fault domains in the ongoing manner.

8. A method as in claim 1, further comprising:

running, on the local computerized device, a first application which writes and modifies shared data stored in the local mirror memory area, the data mirror routine mirroring the shared data from the local mirror memory area to the remote mirror memory area, the shared data which is mirrored by the data mirror routine from the local mirror memory area to the remote mirror memory area being initially protected against inadvertent modification by applications running on the remote computer device while the first application runs on the local computerized device.

9. A method as in claim 8, further comprising:

in response to a failure of the local computerized device, running a second application on the remote computerized device, the second application being constructed and arranged to write and modify the shared data stored in the remote mirror memory area following the failure of the local computerized device.

10. A method as in claim 1, further comprising:

prior to imposing the lock on the shared data, receiving a write operation signal to perform a write operation to write the shared data to the local physical memory, the copy operation being performed in response to the write operation signal; and

only after the lock on the shared data is released, providing an output signal indicating successful completion of the write operation, the lock having been maintained on the shared data until the copy operation was been completed.

11. A method as in claim 10 wherein the page table includes (i) a first set of translations which, for the local non-mirror memory area, maps virtual page addresses to physical page addresses of the local non-mirror memory area, and (ii) a second set of translations which, for the local mirror memory area, maps virtual page addresses to physical extents of the local mirror memory area, each physical extent having an extent size which is less than a size of a page of the local computerized device to provide sparse page shared data storage;

and

wherein performing the copy operation includes moving, as the shared data, data from a particular physical extent having the extent size which is less than the size of a page from the local physical memory to the remote computerized device.

12. A method as in claim 11 wherein observing the page table entry, leaving the data mirror routine inactive, and activating the data mirror routine are performed by a page fault handler which is constructed and arranged to further process page faults occurring in the local non-mirror memory area.

13. A computerized device, comprising:

local physical memory;

an internodal communications interface; and

a controller coupled to the local physical memory and to the internodal communications interface, the controller being constructed and arranged to;

observe a page table entry on the local computerized device, the page table entry being stored in a page table used for managing virtual to physical address translations, tracking page modifications and handling page faults between semiconductor memory and magnetic disk drive memory on the local computerized device,

when the page table entry indicates that shared data corresponding to the page table entry has not been modified on the local computerized device, leave a data mirror routine inactive on the local computerized device, the data mirror routine being configured to copy the shared data from the local physical memory to a remote computerized device, and

when the page table entry indicates that shared data corresponding to the page table entry has been modified on the local computerized device, activate the data mirror routine to mirror the shared data from the local physical memory to the remote computerized device;

wherein the local physical memory of the local computerized device includes a local non-mirror memory area and a local mirror memory area;

wherein remote physical memory of the remote computerized device includes a remote non-mirror memory area and a remote mirror memory area;

wherein the local non-mirror memory area and the remote non-mirror memory area store different data to independently run, on the local and the remote computer devices, different applications performing different tasks over different fault domains in an ongoing manner;

wherein the local mirror memory area and the remote mirror memory area maintain mirroring copies of the shared data to provide fault tolerant redundancy for the shared data across the different fault domains while the local and the remote computer devices independently run the different applications performing the different tasks over the different fault domains in the ongoing manner; and

wherein, in response to activation when the page table entry indicates that shared data corresponding to the page table entry has been modified on the local computerized device, the data mirror routine is constructed and arranged to (i) impose a lock on the shared data, (ii) perform a copy operation which copies the shared data from the local physical memory to the remote computerized device while concurrently maintaining the lock on the shared data, and (iii) release the lock after the copy operation is completed.

14. A computerized device as in claim 13 wherein a computerized system is formed by multiple computerized devices which include the local computerized device and the remote computerized device;

and wherein the controller is further configured to;

pre-define, for each computerized device of the multiple computerized devices forming the computerized system, another computerized device of the multiple computerized devices to operate as a mirror for shared data on that computerized device, the remote computerized device being configured to operate as a mirror for the shared data on the local computerized device.

15. A computerized device as in claim 14 wherein the controller is further configured to:

suffer a failure which (i) prevents the local computerized device from providing access to the shared data on the local physical memory and (ii) enables access to the shared data on the remote computerized device, the shared data on the remote computerized device having been copied to the remote computerized device by the data mirror routine.

16. A computerized device as in claim 14 wherein the controller, when observing the page table entry, is configured to:

inspect a modified bit to determine whether the shared data corresponding to the page table entry has been modified on the local computerized device, the modified bit indicating that the shared data (i) has not been modified when the modified bit is unasserted and (ii) has been modified when the modified bit is asserted.

17. A computerized device as in claim 16 wherein the controller, when activating the data mirror routine to mirror the shared data from the local physical memory to the remote computerized device, is configured to:

in response to a determination that the modified bit has been asserted, mirror the shared data from the local physical memory to the remote computerized device.

18. A computerized device as in claim 17 wherein the computerized system is formed by N computerized devices;

wherein a page of the local physical memory is divided into Q extents;

wherein the shared data resides in a particular extent of the Q extents; and

wherein the controller, when mirroring the shared data from the local physical memory to the remote computerized device, is configured to;

copy the shared data from the particular extent of the Q extents to a corresponding extent in physical memory of the remote computerized device.

19. A computerized device as in claim 13 wherein the local mirror memory area includes (i) a first page storing first data which is updated by a first application running on the local computerized device and (ii) a second page storing second data which is updated by a second application running on the remote computerized device, the first and second applications running concurrently to perform at least some of the different tasks over the different fault domains in the ongoing manner.

20. A computerized device as in claim 13 wherein the controller is further configured to:

run a first application which writes and modifies shared data stored in the local mirror memory area, the data mirror routine mirroring the shared data from the local mirror memory area to the remote mirror memory area, the shared data mirrored by the data mirror routine from the local mirror memory area to the remote mirror memory area being initially protected against inadvertent modification by applications running on the remote computer device while the first application runs on the local computerized device.

21. A computerized device as in claim 13 wherein the controller is further constructed and arranged to:

prior to imposing the lock on the shared data, receive a write operation signal to perform a write operation to write the shared data to the local physical memory, the copy operation being performed in response to the write operation signal, and

only after the lock on the shared data is released, provide an output signal indicating successful completion of the write operation, the lock having been maintained on the shared data until the copy operation was been completed.

22. A computerized device as in claim 21 wherein the page table includes (i) a first set of translations which, for the local non-mirror memory area, maps virtual page addresses to physical page addresses of the local non-mirror memory area, and (ii) a second set of translations which, for the local mirror memory area, maps virtual page addresses to physical extents of the local mirror memory area, each physical extent having an extent size which is less than a size of a page of the local computerized device to provide sparse page shared data storage;

and

wherein performing the copy operation includes moving, as the shared data, data from a particular physical extent having the extent size which is less than the size of a page from the local physical memory to the remote computerized device.