High performance multiple-unit electronic data storage system with checkpoint logs for rapid failure recovery

• US 5,481,694 A
• Filed: 04/24/1992
• Issued: 01/02/1996
• Est. Priority Date: 09/26/1991
• Status: Expired due to Term

• Current Assignee: Samsung Electronics Co. Ltd.
• Original Assignee: Hewlett-Packard Company (HP Inc.)
• Inventors: Stepanov, Alexander A., Wilkes, Andrew J., Chao, Chia-Chiang, English, Robert M., Jacobson, David M.
• Primary Examiner(s): Robertson, David L.
• Application Number:
  US07/873,928
• Time in Patent Office: 1,348 Days
• Field of Search: 395/400, 395/425
• US Class Current: 711/112
• CPC Class Codes: G06F 11/1451 by selection of backup contents
• CPC Class Codes: G06F 2003/0697 device management, e.g. handlers, drivers, I/O schedulers
• CPC Class Codes: G06F 3/0601 Interfaces specially adapted for storage systems
• CPC Class Codes: G11B 20/10 Digital recording or reproducing
• CPC Class Codes: G11B 20/1252 for discontinuous data, e.g. digital information signals, computer programme data
• CPC Class Codes: G11B 2020/1282 in embedded servo fields
• CPC Class Codes: G11B 27/11 by using information not detectable on the record carrier

Abstract

An electronic data storage system including a memory, a plurality of magnetic disk units, and a controller. The memory contains an index cross-referencing logical address with physical addresses, an obsolete list and a free list. In response to a "write " command, the controller selects a physical address according to which segment can be used the most quickly, appends a tag to the data to be written, and writes the data to the selected segment. Appropriate entries are made in the index and the free list. The system recovers from memory loss by using a checkpoint log and a set of checkpoint segments on the disk that together contain backups of the index and other critical information needed to restore the system. Group indices are used for roll-back groups; operations on data in a group are invisible outside the group until after a "commit " command is issued.

Claims

1. An electronic data storage system comprising a memory;

a plurality of storage units each having a plurality of segments for storing blocks of data, each segment identified by a physical address; and

a controller in communication with the memory and the storage units, the controller operative;

to maintain in the memory an index containing entries each having a logical address and a corresponding physical address, an obsolete list containing physical addresses of segments that contain obsolete data, and a free list containing physical addresses of segments that are available to receive data for storage, and

in response to a write command that provides a preselected logical address and an associated block of data for storage, to (a) select a physical address from the free list according to a predefined criterion, (b) obtain a next sequence number from a set of sequential numbers, (c) append to the block a tag that includes the sequence number, the preselected logical address, and an end flag, (d) store the block and its tag in the segment of the selected physical address, (e) delete the selected physical address from the free list, (f) enter the preselected logical address and the selected physical address in the index, (g) delete any previous index entry that contains a logical address which is the same as the preselected logical address, and (h) list the physical address from any such deleted entry in the obsolete list.

2. A system as in claim 1 wherein the controller is operative, in response to a write command that provides a preselected logical address and a plurality of associated blocks of data for storage, for each such block to (a) select a physical address from the free list according to the predefined criterion, (b) obtain a next sequence number from the set of sequential numbers, (c) append a tag that includes the sequence number of that block, the tag for the first block also including the preselected logical address, the tag for each block after the first including a logical address calculated by incrementing the logical address of the preceding block, and the tag for the last block including an end flag, (d) store the block and its tag in the segment of the physical address that was selected for that block, (e) delete the selected physical address from the free list, (f) enter the logical address of the block and the selected physical address in the index, (g) delete any previous index entry that contains a logical address which is the same as the logical address of the block, and (h) list the physical address from any such deleted entry in the obsolete list.

3. A system as in claim 1 wherein the predefined criterion comprises which available segment can be accessed in a minimum amount of time relative to the time that would be required to access others of the available segments.

4. A system as in claim 1 wherein the controller is operative to (a) maintain in the memory a checkpoint log that includes a record of each entry in, and each deletion from, the index, (b) arrange the various index entries in a plurality of checkpoint sections, (c) reserve a plurality of the segments as checkpoint segments, and periodically (d) store the checkpoint log, an image of a sequential one of the checkpoint sections of the memory, and a tag that includes a next sequence number from the set of sequential numbers and an end flag in a sequential one of the checkpoint segments and transfer any physical addresses then in the obsolete list into the free list.

5. A system as in claim 4 wherein, if the contents of the memory are lost, the controller is operative to recover the index and the free list by:

reading the checkpoint segments to obtain the most recent valid segments that collectively contain all of the checkpoint sections of the memory, the recency of a segment being indicated by its sequence number and the validity of a segment being indicated by the presence of an end flag,

starting with the oldest of the obtained checkpoint segments, using the checkpoint sections and checkpoint logs as stored in the checkpoint segments to recover the index up through the most recent checkpoint, and

for all non-checkpoint segments not yet listed in the index, reading the sequence numbers of said segments, entering the physical address of each said segment having a sequence number older than that of the most recent checkpoint segment into the free list, arranging the remaining segments in order of increasing sequence numbers until an invalid segment is detected, entering the logical and physical addresses of each of the arranged segments into the index, and entering the physical address of any remaining segments into the free list.

6. A system as in claim 5 wherein the controller detects an invalid non-checkpoint segment by locating a numerical gap between the sequence numbers of the segments and going back through the segments in reverse order of sequence numbers until a segment having an end-of-block tag is located, the segment following the one having the end-of-block tag being the invalid one.

7. A system as in claim 1 wherein the controller is operative:

to maintain the index as a global index and one or more group indices,

to create a group index in response to a create-group command, each group index containing entries each having the logical address of a block that belongs to the group, a status indicator, and, if the indicated status is "write " or "no-write", the physical address of the block,

to maintain in the memory a reference list that provides a reference count for each physical address that appears either in the global index or in any of the group indices,

to include in the tag appended to a block an identifier of any group to which the block belongs, and

in response to a write command (a) if the command does not include a group identifier, to enter the preselected logical address and the selected physical address in the global index, (b) if the command includes a group identifier, to enter the preselected logical address, the selected physical address and a "write " status indicator in the group index corresponding with the identified group, (c) to increment the reference count for the selected physical address, (d) to delete any previous entry in the same index that contains the same logical address together with a different physical address, (e) to decrement the reference count for the physical address of the deleted entry, (f) if the deleted entry was in the global index, to copy the logical and physical addresses of the deleted entry and a no-write status indicator into each group index, and (g) to increment the reference count for the physical address of the deleted entry once for each group index into which the entry is copied.

8. A system as in claim 7 wherein the controller is operative in response to a delete command that includes a logical address:

if the delete command contains no group identifier, to delete any global index entry that contains that logical address, to decrement the reference count of the physical address in the deleted entry, and to make corresponding entries in the various group indices;

if the delete command contains a group identifier and the logical address is already entered in the corresponding group index, to delete any physical address from the entry, to change the status indicator to "delete", and to decrement the reference count of the deleted physical address, and

if the delete command contains a group identifier and the logical address is not already entered in the corresponding group index, to enter the logical address in the index together with a "delete " status indicator.

9. A system as in claim 8 wherein the controller is operative in response to a commit command that includes a group identifier to delete the corresponding group index and, for each entry in the deleted index:

if the entry has a "no-write " status, to decrement the reference count for the physical address in the entry,

if the entry has a "delete " status, to delete any entry in the global index having the same logical address and to decrement the reference count for the physical address in each such deleted entry, and

if the entry has a "write " status, (a) to copy the logical and physical address of the entry into the global index, (b) to delete any previous entry in the global index that contains the same logical address together with a different physical address, (c) to decrement the reference count for the physical address in any such deleted entry, (d) to copy the logical and physical addresses of any such deleted entry and a "no-write " status indicator into all other group indices, and (e) to increment the reference count for the physical address of the deleted entry once for each group index into which the entry is copied.

10. A system as in claim 9 wherein the controller is operative to enter any delete and commit commands into an operations log and, in response to a sync command, to store the operations log in an available segment and enter the physical address of that segment into the obsolete list.

11. In an electronic data storage system of the kind having a plurality of storage units each having a plurality of segments for storing blocks of data, each segment identified by a physical address, a data storage and retrieval method comprising:

maintaining an index of logical addresses and corresponding physical addresses;

maintaining an obsolete list containing physical addresses of segments that contain obsolete data;

maintaining a free list containing physical addresses of segments that are available to receive data for storage;

selecting a physical address from the free list for storing a block of data;

storing the block at the selected physical address;

removing the selected physical address from the free list;

entering the selected physical address and a predetermined logical address in the index;

deleting any previous index entry that contains a logical address which is the same as the predetermined logical address; and

entering any physical address contained in any such deleted entry in the obsolete list.

12. A method as in claim 11 wherein the step of selecting a physical address from the free list comprises determining which of a plurality of available segments can be accessed in a minimum amount of time relative to the time that would be required to access others of the available segments.

13. A method as in claim 11 and further comprising:

maintaining a checkpoint log that includes a record of each entry in, and each deletion from, the index;

grouping the various index entries in a plurality of checkpoint sections;

storing the checkpoint log and an image of a sequential one of the checkpoint sections in a sequential one of a plurality of predetermined checkpoint segments; and

transfering any physical addresses then in the obsolete list into the free list.

14. A method as in claim 13 and further comprising:

appending a next sequential one of a set of sequence numbers to each block of data when such block is stored; and

appending a next sequential one of the set of sequence numbers to each checkpoint section when such checkpoint section is stored.

15. A method as in claim 14 and further comprising recovering a lost index and free list by:

reading the checkpoint segments to obtain the most recent valid segments that collectively contain all of the checkpoint sections;

recovering the index up through the most recent checkpoint segment by means of the checkpoint sections and checkpoint logs stored in the checkpoint segments;

reading the sequence numbers of any non-checkpoint segments not in the recovered index;

entering the physical address of each said segment having a sequence number lower than that of the most recent checkpoint segment into the free list;

arranging those segments having sequence numbers higher than that of the most recent checkpoint segment in order of increasing sequence numbers until a gap occurs between the sequence numbers of two such segments;

entering the logical and physical addresses of each of the arranged segments having a sequence number below the gap into the index; and

entering the physical address of each of the arranged segments having a sequence number above the gap into the free list.