Recording medium storing data allocation control program, data allocation control device, data allocation control method, and multi-node storage-system

A data allocation control device for managing data allocation when data is distributively stored in a plurality of disk nodes which are shifted to a power saving mode if no access is performed for a certain time, the device comprising:

• an access statistical information collecting unit for collecting access statistical information representing the number of accesses performed to respective data managed by the disk nodes within a given time slot, from an access statistical information storage unit which stores the access statistical information;
• a no-access probability calculating unit for referring to the access statistical information and setting, as a no-access-to-data probability for each data, a value that is increased for data with a smaller number of accesses within the given time slot;
• an allocation pattern candidate producing unit for producing a plurality of allocation pattern candidates each indicating the disk nodes in which the respective data are to be stored;
• an expectation calculating unit for calculating, for each of the produced allocation pattern candidates, the no-access-to-data probabilities for the respective data allocated in each of the disk nodes when the respective data are allocated according to the allocation pattern candidate, to thereby calculate a no-access-to-node probability representing a probability that access is not performed to each disk node, and tallying the respective no-access-to-node probabilities for the disk nodes, thus obtaining a no-access period expectation that represents an expected value of occurrence of a no-access period during which access is not performed to some of the disk nodes;
• an allocation pattern deciding unit for selecting, as an allocation pattern for data reallocation, one of the plurality of produced allocation pattern candidates with the largest one of the calculated no-access period expectations;
• and
• a data reallocation instructing unit for instructing the disk nodes to reallocate the respective data in accordance with the allocation pattern selected by the allocation pattern deciding unit.

A data allocation control method for managing data allocation by a computer when data is distributively stored in a plurality of disk nodes which are shifted to a power saving mode if no access is performed for a certain time, the method comprising:

• collecting access statistical information representing the number of accesses performed to respective data managed by the disk nodes within a given time slot from an access statistical information storage unit which stores the access statistical information;
• referring to the access statistical information and setting, as a no-access-to-data probability for each data, a value that is increased for data with a smaller number of accesses within the given time slot;
• producing a plurality of allocation pattern candidates each indicating the disk nodes in which the respective data are to be stored;
• calculating, for each of the produced allocation pattern candidates, the no-access-to-data probabilities for the respective data allocated in each of the disk nodes when the respective data are allocated according to the allocation pattern candidate, to thereby calculate a no-access-to-node probability representing a probability that access is not performed to each disk node, and tallying the respective no-access-to-node probabilities for the disk nodes, thus obtaining a no-access period expectation that represents an expected value of occurrence of a no-access period during which access is not performed to some of the disk nodes;
• selecting, as an allocation pattern for data reallocation, one of the plurality of produced allocation pattern candidates with the largest one of the calculated no-access period expectations;
• and
• instructing the disk nodes to reallocate the respective data in accordance with the selected allocation pattern.

A multi-node storage system for managing data on a network in a distributed manner, the system comprising:

• storage devices for storing respective data;
• a plurality of disk nodes;
• and
• each of the disk nodes comprising:
• an access statistical information storage unit for storing access statistical information representing the number of accesses performed to the data stored in corresponding one of the storage devices within a given time slot;
• an access statistical information update unit for updating the access statistical information whenever access is performed to the data stored in the storage device within the given time slot;
• a data reallocation unit for, upon receiving an instruction for data reallocation, transferring the data stored in the storage devices in accordance with the instruction;
• and
• an operation mode control unit for, when no access is performed to the data stored in the storage device for a certain time, shifting the disk node to power saving mode in which at least a part of functions of the disk node is stopped,
• a control node, the control node comprising:
• an access statistical information collecting unit for collecting the access statistical information from the access statistical information storage unit in each of the disk nodes;
• a no-access probability calculating unit for referring to the access statistical information and setting, as a no-access-to-data probability for each data, a value that is increased for data with a smaller number of accesses within the given time slot;
• an allocation pattern candidate producing unit for producing a plurality of allocation pattern candidates each indicating the disk nodes in which the respective data are to be stored;
• an expectation calculating unit for calculating, for each of the produced allocation pattern candidates, the no-access-to-data probabilities for the respective data allocated in each of the disk nodes when the respective data are allocated according to the allocation pattern candidate, to thereby calculate a no-access-to-node probability representing a probability that access is not performed to each disk node, and tallying the respective no-access-to-node probabilities for the disk nodes, thus obtaining a no-access period expectation that represents an expected value of occurrence of a no-access period during which access is not performed to some of the disk nodes;
• an allocation pattern deciding unit for selecting, as an allocation pattern for data reallocation, one of the plurality of produced allocation pattern candidates with the largest one of the calculated no-access period expectations;
• and
• a data reallocation instructing unit for instructing the disk nodes to reallocate the respective data in accordance with the allocation pattern selected by the allocation pattern deciding unit.