Back to results
US 8,527,454 B2Granted
Data replication using a shared resource
AssigneeEMC Corporation
InventorDhairesh Oza
Life of this patent
Estimating the term…
Abstract
Techniques are presented for data replication. Two or more nodes are engaged in data replication with one another. When a node acquires a write lock, the write is processed to a local disk associated with the node and communicated to a shared disk. During recovery of a particular node following a failure, the shared disk is accessed to assist in synchronizing that particular node.
Claims23
A method, comprising:
- acquiring a write lock on a first node;
- writing blocks associated with the write lock to one or more local disks of the first node;
- simultaneously writing the blocks associated with the write lock to a shared disk over a wide-area network connection, the shared disk is shared with a second node and the first node, and the first node and second node are replicated with one another via the shared disk that is accessed by the first node and the second node over the wide-area network connection, the shared disk is on a particular node that is independent of the first node and the second node, the blocks associated with the write lock are also written to the second node, furthermore writes to the shared disk, the first node and the second node occur when there are no failures, each write operation made to either the first node or the second node results in N+1 writes where N represents an integer for a total number of nodes replicated and the addition of one represents the shared node;
- providing profile information that identifies an alternative shared disk for automatic usage when the shared disk becomes unresponsive and the alternative shared disk is also replicated with the shared disk;
- and
- releasing the write lock after the writing of the blocks to the one or more local disks of the first node and the writing of the blocks to the shared disk have completed.
3 more elements
A method, comprising:
- monitoring progress of synchronization between two or more nodes that are replicas and synchronized with one another;
- detecting a failure in a particular node while a write was attempting to take place but does not complete, and the write was being attempted to a shared disk associated with the two or more nodes, the shared disk on a different node that is independent of the particular node and the two or more nodes and the shared disk accessible over a wide-area network connection to the particular node and the two or more nodes, the shared disk is being used for managing multi-node replication for the particular node and the two or more nodes and the shared disk is independent of the particular node and the two or more nodes, and profile information identifies an alternative shared disk to the shared disk for use when the shared disk fails and the alternative shared disk is replicated with the shared disk;
- and
- clearing a log in the shared disk associated with a partial write associated with the attempted write to ensure that there is no indication of the attempted write in the log, the two or more nodes including the particular node communicate with the shared disk and the particular node knows when the write fails when the particular node recovers because the particular node will not see any indication of the write when it comes back online based on the cleared log missing any indication of the write.
1 more element
A system, comprising:
- a replication service implemented in a machine-accessible and readable medium and to process on a machine associated with a first node;
- and
- a shared disk implemented as a device accessible to the first node over a wide-area network connection and locally accessible to a different machine associated with a second node, the second node is independent of the first node and a third node;
- the replication service is to keep a first local disk associated with the first node synchronized with a third local disk associated with the third node by acquiring write locks for writes and performing the writes against the first local disk and communicating the writes to the shared disk via the wide-area network connection and then releasing the write locks, the shared disk including tracking and audit information that also includes identity information for resources that make the writes on the first local disk and the third local disk, each write operation performed against the first local disk, the shared disk, and the third local disk, the shared disk for managing replication between the first local disk and the third local disk, the shared disk replicated with an alternative shared disk and the alternative shared disk identified via profile information.
1 more element
A system, comprising:
- a monitor service implemented in a machine-accessible and readable medium and to process on a machine associated with a first node;
- and
- a shared disk implemented as a device accessible to the first node over a wide-area network connection and accessible to at least one additional node, the at least one addition node is independent of the first node and a particular node;
- the monitor service is to dynamically and in real time keep track of replication services that process on the at least one additional node and writes information to the shared disk when the particular node fails that permits the particular node to synchronize with the at least one additional node when the particular node recovers from a failure situation, the particular node and the at least one additional node dynamically communicate with the monitor service and the monitor service keeps the shared disked updated with the write information, and wherein each write operation is performed on the particular node, the at least one additional node, and the shared disk, profile information provides identification to the monitor service of an alternative shared disk for usage when the shared disk fails and the alternative shared disk is replicated with the shared disk.
1 more element
Dependencies are inferred from the claim text.