Using virtual disks for disk system checkpointing

Claim 1

A method of checkpointing lost data in a multiprocessor, distributed disk array computer system to protect against losing data, comprising: providing a logical disk representing the data stored on one or more physical disks associated with each processor of the multiprocessor system; assigning to each of the processors a set of virtual disks, the virtual disks being mapped via the logical disk to data storage locations on the physical disks; designating one virtual disk in each set to be a user virtual disk, and updating the data stored in the user virtual disk when desired; designating another of the virtual disks in each set to be a checkpoint virtual disk, the checkpoint virtual disks containing most recent checkpoint data and parity blocks for reconstructing the checkpoint data in the event of a physical disk failure; designating another of the virtual disks in each set to be a constructing checkpoint virtual disk, the constructing checkpoint virtual disk containing data desired to be saved as checkpoint data; constructing new parity blocks derived from the data desired to be saved as checkpoint data in the constructing checkpoint virtual disks; when the new parity blocks have been constructed, assigning the constructing checkpoint virtual disks to be new checkpoint virtual disks, and assigning the checkpoint virtual disks to be spare virtual disks, or saving them as a finite sequence of checkpoints in additional virtual disks.

Claim 2

A method of checkpointing lost data in a multiprocessor, distributed disk array computer system to protect against losing data, comprising: providing a logical disk representing the data stored on one or more physical disks associated with each processor of the multiprocessor system; assigning to each of the processors a set of virtual disks, the virtual disks being mapped via the logical disk to data storage locations on the physical disks; designating one virtual disk in each set to be a user virtual disk, and updating the data stored in the user virtual disk when desired; designating another of the virtual disks in each set to be a checkpoint virtual disk, the checkpoint virtual disks containing most recent checkpoint data and parity blocks for reconstructing the checkpoint data in the event of a physical disk failure; designating another of the virtual disks in each set to be a constructing checkpoint virtual disk, the constructing checkpoint virtual disk containing data desired to be saved as checkpoint data; constructing new parity blocks derived from the data desired to be saved as checkpoint data in the constructing checkpoint virtual disks; when the new parity blocks have been constructed, assigning the constructing checkpoint virtual disks to be new checkpoint virtual disks, and assigning the checkpoint virtual disks to be spare virtual disks, or saving them as a finite sequence of checkpoints in additional virtual disks.

Claim 6

A method of checkpointing data stored on one or more storage devices in a computer employing a plurality of processors and a plurality of distributed disk storage devices, comprising the steps of: classifying all storage space associated with the processors as one or more logical disks; assigning a plurality of virtual disks to a processor, the virtual disks together comprising a subset of the logical disks; assigning functions to the virtual disks of each processor such that each function is assigned to one or more virtual disks of each processor; storing a checkpoint of the distributed disk system in one or more virtual disks, portions of the checkpoint being distributed among one or more of the virtual disks, wherein the checkpoint represents the state of the distributed disk system at a particular time; periodically updating the checkpoint.

Claim 7

A method of checkpointing data stored on one or more storage devices in a computer employing a plurality of processors and a plurality of distributed disk storage devices, comprising the steps of: classifying all storage space associated with the processors as one or more logical disks; assigning a plurality of virtual disks to a processor, the virtual disks together comprising a subset of the logical disks; assigning functions to the virtual disks of each processor such that each function is assigned to one or more virtual disks of each processor; storing a checkpoint of the distributed disk system in one or more virtual disks, portions of the checkpoint being distributed among one or more of the virtual disks, wherein the checkpoint represents the state of the distributed disk system at a particular time; periodically updating the checkpoint.

Claim 13

A method of constructing a checkpoint in a multiprocessor system, comprising: maintaining a first block list corresponding to a group of blocks stored in memory for an active version of a file; creating a second block list, identical to the first block list, and maintaining the second block list as a checkpoint; modifying data in the first group of blocks, comprising the steps of: copying data residing in a first disk location to a new disk location; overwriting some or all of the data in the new disk location with new data; replacing the entry in the first block list which corresponds to the first disk location with the address of the new disk location; accounting for the number of blocks addressed by the first and second block lists, comprising the steps of: providing used and reserved block counters, and a maximum limit for the processor; for each file, providing a new block counter and a reserved block counter; setting the used block counter of each file to be the union of blocks used by the checkpoint and user block lists; setting the reserved block counter of each file to be equal to the number of blocks in the checkpoint block list that are not also in the user block list of the file; when a block of the user block list is overwritten, incrementing the used block counter of that file and of the system by one and decrementing the reserved block counter of that file and of the system; setting the total number of used blocks to be equal to the sum of the used block counts for each file; setting the total number of reserved blocks to equal the sum of reserved block counts of each file; maintaining the counters when files are deleted, overwritten, resized or checkpointed.

Claim 14

A method of constructing a checkpoint in a multiprocessor system, comprising: maintaining a first block list corresponding to a group of blocks stored in memory for an active version of a file; creating a second block list, identical to the first block list, and maintaining the second block list as a checkpoint; modifying data in the first group of blocks, comprising the steps of: copying data residing in a first disk location to a new disk location; overwriting some or all of the data in the new disk location with new data; replacing the entry in the first block list which corresponds to the first disk location with the address of the new disk location; accounting for the number of blocks addressed by the first and second block lists, comprising the steps of: providing used and reserved block counters, and a maximum limit for the processor; for each file, providing a new block counter and a reserved block counter; setting the used block counter of each file to be the union of blocks used by the checkpoint and user block lists; setting the reserved block counter of each file to be equal to the number of blocks in the checkpoint block list that are not also in the user block list of the file; when a block of the user block list is overwritten, incrementing the used block counter of that file and of the system by one and decrementing the reserved block counter of that file and of the system; setting the total number of used blocks to be equal to the sum of the used block counts for each file; setting the total number of reserved blocks to equal the sum of reserved block counts of each file; maintaining the counters when files are deleted, overwritten, resized or checkpointed.