Checkpointing shared state in distributed systems
AssigneeAmazon Technologies, Inc.
InventorsAllan Henry Vermeulen, Timothy Andrew Rath
Life of this patent
Estimating the term…
Abstract
To checkpoint a shared state in a distributed system, a node may first isolate itself from communication with other nodes of a distributed system. The node may then store a checkpoint of the shared state. The node then restarts and attempts to initialize its operating state from the stored checkpoint. In response to successfully initializing its operating state, the node restores communication with one or more other nodes of the distributed system. The node then indicates to the one or more other nodes that the stored checkpoint of the shared state is valid.
Claims20
A computer-implemented method, comprising:
- isolating a given node from communication with other nodes of a distributed system, in response to a determination to record a checkpoint of a shared state of the distributed system, wherein the given node isolates itself from communication with the other nodes of the distributed system, wherein the given node and the other nodes of the distributed system are implemented via one or more hardware computing devices;
- storing, by the isolated given node, the checkpoint of the shared state;
- restarting and attempting, by the isolated given node, to initialize its operating state from the stored checkpoint;
- in response to successfully initializing its operating state from the stored checkpoint:
- restoring, by the given node, communication with one or more of the other nodes of the distributed system;
- and
- indicating to the one or more others of the nodes of the distributed system that the stored checkpoint is valid.
A non-transitory computer-readable storage medium storing program instructions that when executed by a hardware computer system cause the computer system to implement:
- a given node of a distributed system isolating itself from communication with other nodes of a plurality of nodes of the distributed system;
- the isolated given node storing a checkpoint of a shared state of the distributed system;
- the given node restarting and attempting to initialize its operating state from the stored checkpoint;
- and
- the given node is configured to respond to successfully initialize its operating state from the stored checkpoint, by:
- restoring communication with one or more others of the plurality of nodes;
- and
- indicating to the one or more others of the plurality of nodes that the stored checkpoint is valid.
A computer, comprising:
- one or more hardware processors and memory storing program instructions that when executed by the one or more hardware processors implement a given node of a distributed system, wherein the given node is configured to respond to being chosen to record a checkpoint of a shared state of the distributed system, wherein to respond the given node is further configured to:
- isolate itself from communication with other nodes of the distributed system;
- store the checkpoint of the shared state;
- restart and attempt to initialize its operating state from the stored checkpoint;
- and
- responsive to successful initialization of its operating state from the stored checkpoint:
- restore communication with one or more others of the nodes of the distributed system;
- and
- indicate to the one or more others of the nodes that the stored checkpoint is valid.
Dependencies are inferred from the claim text.