Checkpoint service

A method comprising:

• an application storing a checkpoint segment of the application's state to a local storage of a first node on which the application is executing, wherein the local storage is directly accessible only to the first node;
• the application registering the checkpoint segment with a checkpoint service that is separate from the application;
• the checkpoint service replicating a checkpoint segment from the first local storage of the first node to at least one other node responsive to the registering and the storing;
• and
• loading a copy of the replicated checkpoint segment from the at least one other node into a second local storage of a second node responsive to a request from the second node to load the copy, wherein the second node is to execute the application, wherein the checkpoint segment is identified by a checkpoint segment name that is unique within a cluster including the first node and the second node.

A cluster comprising a plurality of nodes, wherein each of the plurality of nodes comprises a computer, and

• wherein a checkpoint service executing on a first node of the plurality of nodes is configured to replicate a checkpoint segment from a first local storage of the first node to at least one other node of the plurality of nodes,
• wherein the checkpoint segment is stored into the first local storage by an application executing on the first node separate from the checkpoint service, and
• wherein the application is configured to register the checkpoint segment with the checkpoint service and wherein the checkpoint service is configured to replicate the checkpoint segment responsive to the application registering the checkpoint segment and storing the checkpoint segment to the local storage, wherein the local storage is directly accessible only to the first node of the plurality of nodes, and
• wherein the checkpoint segment comprises a state of the application, and
• wherein a checkpoint service executing on a second node of the plurality of nodes is configured to load a copy of the replicated checkpoint segment from the at least one other node into a second local storage of the second node responsive to a request to load the copy, wherein the second node is to execute the application, wherein the checkpoint segment is identified by a checkpoint segment name that is unique within a cluster including the first node and the second node.