Data management system and method for processing distributed data
AssigneesSK hynix Inc., Seoul National University R&Db Foundation
InventorsDaedong Park, Seongsoo Hong
Life of this patent
Estimating the term…
Abstract
A data management system and method for processing distributed data, which adaptively adjusts the amounts of data to be read from a plurality of storages, depending on I/O performances of the respective storages. The data management system includes a plurality of storage nodes each storing duplicate data of original data, and a distributed data management unit providing read requests for parts of the duplicate data to the respective storage nodes, combining the read parts into a complete version of the data, and outputting the complete version of the data.
Claims14
A data management system comprising:
- a plurality of storage nodes each suitable for storing duplicate data of original data;
- and
- a distributed data management unit suitable for providing read requests for parts of the duplicate data to the respective storage nodes, combining the read parts into a complete version of the original data, and outputting the complete version of the original data,
- wherein the distributed data management unit comprises:
- an input/output (I/O) performance monitoring unit suitable for monitoring I/O performances of the respective storage nodes by comparing current bandwidth capacities for the respective storage nodes with an allocated bandwidth of a corresponding one among the plurality of storage nodes;
- and
- an I/O bandwidth setting unit suitable for determining and allocating I/O bandwidths for the respective storage nodes by referring to the I/O performances monitored the I/O performance monitoring unit by adjusting the allocated I/O bandwidths for the respective storage nodes when the current bandwidth capacities of one or more of the plurality of storage nodes are smaller than the allocated bandwidth of the one or more of the plurality of storage nodes.
A data management method comprising:
- allocating a bandwidth to each of a plurality of storage nodes;
- providing read requests for parts of a duplicate data stored in the respective storage nodes to the respective storage nodes;
- combining the parts of the duplicate data read from the respective storage nodes into a complete version of data, and outputting the complete version of the data;
- monitoring input/output (I/O) performances of the respective storage nodes;
- and
- adjusting allocated bandwidths for the respective storage nodes according to the monitored I/O performances,
- wherein the monitoring of the I/O performances comprises comparing current bandwidth capacities for the respective storage nodes with allocated bandwidth of a corresponding one among the plurality of storage nodes, and
- wherein the adjusting of the bandwidths comprises adjusting the allocated bandwidths of the respective storage nodes when the current bandwidth capacities of one or more of the plurality of storage nodes are smaller than the allocated bandwidth of the one or more of the plurality of storage nodes.
Dependencies are inferred from the claim text.