Distributed system having an improved method and apparatus for checkpoint taking

A node apparatus for a distributed system having a storage apparatus connected thereto, the node apparatus operating in accordance with a server-client model, the node apparatus comprising:

• a client comprising:
• checkpoint taking means for, during normal operation of the distributed system, taking a checkpoint regarding a resource managed by a backup server;
• storage means for storing the checkpoint taken by said checkpoint taking means;
• means for storing the checkpoint taken by said checkpoint taking means into the storage apparatus;
• and
• means for sending out information regarding the resource from the client to the backup server based on the checkpoint stored in said storage means so as to allow, when another node apparatus crashes, the backup server to collect information regarding the resource to perform recovery of the server.

A node apparatus for a distributed system having a storage apparatus connected thereto, the node apparatus operating in accordance with a server-client model, the node apparatus comprising:

• a client comprising:
• checkpoint taking means for, during normal operation of the distributed system, taking a checkpoint regarding a resource managed by a primary resource managing server having an associated backup resource managing server in another node on the distributed system;
• and
• means for storing the checkpoint taken by said checkpoint taking means in the storage apparatus.

A node apparatus for a distributed system having a plurality of other node apparatus and a storage apparatus connected thereto, the node apparatus operating in accordance with a server-client model, the node apparatus comprising:

• a client comprising:
• checkpoint taking means for, during normal operation of the distributed system, taking a checkpoint regarding a resource managed by a primary resource managing server having an associated backup server on the distributed system;
• storage means for storing the checkpoint taken by said checkpoint taking means;
• means for storing the checkpoint taken by said checkpoint taking means on the storage apparatus;
• and
• means for sending out information regarding resources managed by the primary resource managing server from the client to the corresponding backup resource managing server based on the checkpoint stored in said storage means so as to allow, when one of plurality of other node apparatus having a primary resource managing server crashes and the node apparatus receives an inquiry from said backup resource managing server for information regarding the resources managed by the crashed primary resource managing server, said backup resource managing server to re-construct an internal condition of said backup resource managing server to perform recovery of the server.

A node apparatus for a distributed system having a plurality of other node apparatuses and a storage apparatus connected thereto, the node apparatus operating in accordance with a server-client model, the node apparatus comprising:

• a client having an associated backup client in one of the other node apparatus in the distributed system said client comprising:
• checkpoint taking means for, during normal operation of the distributed system, taking a checkpoint regarding a resource managed by a primary resource managing server having a backup managing server;
• storage means for storing the checkpoint taken by said checkpoint taking means;
• means for storing the checkpoint taken by said checkpoint taking means into the storage apparatus;
• means for reading, when a node apparatus other than the node apparatus having the backup client and the primary resource managing server crashes, a checkpoint regarding the client in the crashing primary resource managing server from the storage apparatus to allow the backup client to recover a condition regarding the resources of the server;
• notification means for notifying the backup resource managing server of the recovery;
• and
• means for sending information regarding a resource to the backup resource managing server based on the checkpoint stored so as to allow, when the node apparatus receives an inquiry from the backup resource managing server for information regarding the resources managed by the resource managing server after the notification of the recovery is notified by said notification means, the backup resource managing server to collect information and re-construct an internal condition of the backup resource managing server based on the thus collected information to perform recovery of the server.

A node apparatus for a distributed system having a storage apparatus connected thereto node apparatus operating in accordance with a server-client model, the node apparatus comprising:

• a client comprising:
• checkpoint taking means for, during normal operation of the distributed system, taking a checkpoint regarding a token for permitting the client to use resources managed by a server;
• and
• means for storing the checkpoint regarding the token taken into the storage apparatus.

A node apparatus for a distributed system having a storage apparatus connected thereto, the node apparatus operating in accordance with a server-client model, the node apparatus comprising:

• a client comprising:
• checkpoint taking means for, during normal operation of the distributed system, taking a checkpoint regarding a token for permitting the client to use resources managed by a backup resource managing server;
• storage means for storing information regarding the token taken by said checkpoint taking means;
• means for storing the checkpoint regarding the token into the storage apparatus;
• and
• means for sending out information regarding the token from the client to the backup resource managing server based on the checkpoint stored in the means for storage allowing, when another node apparatus having another resource manager therein crashes, the backup resource managing server to collect information regarding the token by way of the client to perform recovery of the backup resource managing server.

A node apparatus for a distributed system having a storage apparatus, the node apparatus operating in accordance with a server-client model, the node apparatus comprising:

• a client comprising:
• checkpoint taking means for, during normal operation of the distributed system, taking a checkpoint regarding a token for permitting the client to use resources managed by a primary resource managing server;
• storage means for storing information regarding the token;
• means for storing the checkpoint regarding the token into the storage apparatus;
• and
• means for sending out information regarding the token from the client to a backup resource managing server based on the checkpoint regarding the token so as to allow, when another one of a plurality of node apparatus connected to the distributed system, and having a primary resource managing server, crashes, the backup resource managing server to collect information regarding the token by way of the client to perform recovery of the primary resource managing server.

A recovery method for a server contained in one of a plurality of node apparatus connected in a distributed system comprising the steps of:

• taking by the client, during normal operation of the distributed system, checkpoints regarding resources managed by the server in one of the node apparatus while the server does not take the checkpoints regarding the resources;
• storing the checkpoints into an internal storage provided in a client in one of the node apparatus and into a storage apparatus connected to the distributed system;
• collecting, when one of the plurality of node apparatus crashes, information regarding the resources based on the checkpoints;
• and
• performing recovery of the server based on the collected information.

A recovery method for primary resource managing server contained in one of a plurality of node apparatus in a distributed system having a backup resource managing server contained in different one of the plurality of node apparatus comprising the steps of:

• taking by the client, during normal operation of the distributed system, checkpoints regarding resources managed by the primary resource managing server, the primary resource managing server not taking the checkpoints, and storing the taken checkpoints into an internal storage means in a client and a storage apparatus connected to the distributed network;
• when the primary resource managing server crashes, asking the client for information regarding the resources managed by the primary resource managing server based on the checkpoints stored in the client or the storage apparatus to collect information for the backup resource managing server;
• and
• re-constructing an internal condition of the backup resource managing server based on the collected information to perform recovery of the server.

A recovery method for a primary resource managing server contained in one of a plurality of node apparatus connected in a distributed system having a backup resource managing server, the method comprising the steps of:

• taking by the client, during normal operation of said distributed system, checkpoints regarding resources managed by the primary resource managing server while the primary resource managing server does not take such checkpoints and storing the checkpoints into a storage provided in a client in one of the node apparatus and into a storage apparatus connected to the distributed system;
• reading, when one of the plurality of node apparatus containing the client crashes, a checkpoint regarding a client present in the node apparatus which has crashed from the storage apparatus;
• notifying, at a point of time when a condition regarding resources of the server recovers, the backup resource managing server of such recovery;
• and
• asking for, upon reception of the notification of the recovery by the backup resource managing server from the recovering client, the client for information regarding the resources managed by the primary resource managing server to collect information from the storage in the client and re-constructing an internal condition of said backup resource managing server based on the collected information to perform recovery of the primary resource managing server.

A recovery method for a server contained in one of a plurality of node apparatus contained in a distributed system, the method comprising the steps of:

• taking by the client, during normal operation of the distributed system, checkpoints regarding tokens for permitting a client contained in one of the node apparatus to use resources managed by the server while the server does not take the checkpoints regarding the tokens;
• storing the checkpoints into a storage in the client and a storage apparatus connected to the distributed system;
• and
• collecting, when one of the plurality of node apparatus crashes, information regarding the tokens based on the check points;
• and
• performing recovery of the server based on the collected information.

A recovery method for a primary resource managing server contained in one of a plurality of node apparatus connected in a distributed system having a backup resource managing server distributed in different node apparatus, the method comprising the steps of:

• taking by the client, during normal operation of said distributed system, checkpoint regarding tokens for permitting a client contained in one of the node apparatus to use resources managed by the primary resource managing server while the primary resource managing server does not take the checkpoints regarding the tokens;
• storing the checkpoints regarding the tokens into a storage in the client and into a storage apparatus connected to the distributed system;
• and
• when one of said plurality of node apparatus in which said primary resource managing server is present crashes a backup resource managing server, present in a different one of said plurality of node apparatus from the node apparatus in which said primary resource managing server is present inquires, based on the information regarding the tokens stored in one of said storage means and said storage apparatus in the ordinary operation of said distributed system, the client for information regarding the tokens to collect information regarding the tokens and re-constructing an internal condition of said backup resource managing server based on the thus collected information to perform recovery of the server.