Relocating paged metadata
View Patent ↗A storage array uses paged metadata. Each storage director has access to a plurality of object storage systems which describe locations of paged metadata in backing storage. Each object storage system includes different types of inodes which describe objects in backing storage. The object storage systems are used to locate and relocate metadata for loading into global memory, and creation and deletion of objects. An object storage system may be selected based on factors including ratio of different inode types, locality of object usage and anticipated object activity level.
1. An apparatus comprising:
a plurality of storage directors that utilize one or more physical non-volatile storage devices to present a logical storage device to at least one host, and which temporarily copy pages of metadata from the non-volatile storage devices into a shared memory as needed to service input-output commands from the at least one host, each of the pages having an identical fixed-size and indicating locations of host data on the non-volatile storage devices, each of the storage directors having access to a plurality of object storage systems in the shared memory that is accessible to each of the storage directors,
wherein each object storage system is uniquely associated with a corresponding same-fixed-size portion of the non-volatile storage devices,
wherein each object storage system comprises one or more sets of inodes with pointers to locations of corresponding pages of metadata in the same-fixed-size portion of the non-volatile storage devices associated with that object storage system,
wherein one of the object storage systems is selected as a destination object storage system by a first storage director of the plurality of storage directors to host an object associated with pages of metadata based on sequential space availability in the shared memory to maintain an inode set that describes the locations of the pages of the metadata of the object and sequential page availability in the associated same-fixed-size portion of the non-volatile storage devices to store the object, and
wherein the first storage director relocates the object from an origin object storage system to the destination object storage system.
2. The apparatus of claim 1 wherein the first storage director locks the origin object storage system, thereby preventing addition, deletion, and change to the inode set in preparation for relocating the object.
3. The apparatus of claim 2 wherein the first storage director selects one inode of the inode set at a time for relocation.
4. The apparatus of claim 3 wherein the inode selected for relocation is Read and Write locked.
5. The apparatus of claim 4 wherein the Read and Write lock is released after the inode is relocated.
6. The apparatus of claim 5 wherein the origin object storage system is unlocked after the object is relocated.
7. The apparatus of claim 1 wherein pages are allocated to inodes in the destination object storage system.
8. The apparatus of claim 1 wherein inodes are added or deleted to the inode set in the destination object storage system based on object size change.
9. The apparatus of claim 1 wherein a red black tree of each storage director is updated.
10. A method comprising:
in a storage array having a plurality of storage directors that utilize one or more physical non-volatile storage devices to present a logical storage device to at least one host, and which temporarily copy pages of metadata from the non-volatile storage devices into a shared memory as needed to service input-output commands from the at least one host, each of the pages having an identical fixed-size and indicating locations of host data on the non-volatile storage devices, each of the storage directors having access to a plurality of object storage systems in the shared memory that is accessible to each of the all storage directors, the object storage systems comprising sets of inodes with pointers to locations of corresponding pages of metadata in at least one of the non-volatile storage devices, wherein each object storage system is uniquely associated with a corresponding same-fixed-size portion of the non-volatile storage devices relocating an object associated with paged metadata by:
calculating a number of sequential pages required to store the paged metadata of the object on one of the same-fixed-size portions of the non-volatile storage devices;
calculating required sequential space to maintain an inode set that describes the locations of the calculated number of sequential pages in the shared memory;
selecting one of the object storage systems as a destination object storage system based on sequential space availability in the shared memory to accommodate the inode set and sequential page availability in the associated same-fixed-size portion of the non-volatile storage devices to accommodate the object; and
relocating the object from an origin object storage system to the selected destination object storage system.
11. The method of claim 10 comprising locking the origin object storage system and thereby preventing addition, deletion, and change to the inode set in preparation for relocating the object.
12. The method of claim 11 comprising selecting one inode of the inode set at a time for relocation.
13. The method of claim 12 comprising Read and Write locking the inode selected for relocation.
14. The method of claim 13 comprising releasing the Read and Write lock after the inode is relocated.
15. The method of claim 14 comprising unlocking the origin object storage system after the object is relocated.
16. The method of claim 10 comprising allocating pages to inodes in the destination object storage system.
17. The method of claim 10 comprising adding or deleting inodes in the destination object storage system based on object size change.
18. The method of claim 10 comprising updating a red black tree of each storage director.
19. The method of claim 10 comprising releasing the object from the origin object storage system.
20. A method comprising:
in a storage array having a plurality of storage directors that utilize one or more physical non-volatile storage devices to present a logical storage device to at least one host, and which temporarily copy pages of metadata from the non-volatile storage devices into a shared memory as needed to service input-output commands from the at least one host, each of the pages having an identical fixed-size and indicating locations of host data on the non-volatile storage devices, each of the storage directors having access to a plurality of object storage systems in the shared memory that is accessible each of the storage directors, the object storage systems comprising inodes with pointers to locations of corresponding pages of metadata in at least one of the non-volatile storage devices, wherein each object storage system is uniquely associated with a corresponding same-fixed-size portion of the non-volatile storage devices, relocating an object associated with paged metadata by:
calculating a number of sequential pages required for to store the paged metadata of the object;
calculating required sequential space to maintain an inode set that describes the locations of the calculated number of sequential pages in the shared memory;
locking an origin object storage system that hosts the object;
selecting one of the object storage systems as a destination object storage system for the object based on sequential space availability in the shared memory to accommodate the inode set and sequential page availability in the associated same-fixed-size portion of the non-volatile storage devices to accommodate the object;
allocating pages to inodes in the destination object storage system;
selecting one inode of the inode set at a time for relocation;
Read and Write locking the inode selected for relocation;
copying the selected inode from the origin object storage system to the destination object storage system;
releasing the Read and Write lock after the selected inode is relocated; and
the origin object storage system after the object is relocated.