Distributed storage system and data migration method
A highly accurate data migration pattern is decided on the bases of usages of a plurality of kinds of resources and by taking influence of execution of data migration on each resource into consideration. In a distributed storage system, a monitor acquires a per-period usage of each of the plurality of kinds of the resources per data, a performance simulator estimates a per-period usage of each resource on a migration destination node when the data is migrated by using the acquired per-data resource usage and calculates an estimate of a latency from the estimated usage of each resource and an optimizer decides a migration pattern of the data on the basis of the estimate of the latency.
1. A distributed storage system comprising:
a plurality of nodes which are connected with one another over a network, have processors and memories individually and thereby provide a volume into/from which a host system inputs/outputs data; and
a storage medium which stores the data which is input into/output from the volume, wherein
the processor of any one of the nodes
acquires a per-period usage of each of a plurality of kinds of resources per data,
estimates a per-period usage of each resource on a migration destination node when the data is migrated to the migration destination node by using the acquired per-data resource usage and calculates an estimate of a latency from the estimated usage of each resource, and
decides a migration pattern of the data on the basis of the estimate of the latency;
wherein the processor acquires the input/output together with characteristics of the input/output, and calculates a usage of each resource by the input/output on the basis of the characteristics of the input/output;
wherein types of the input/output, that is, read input/output and write input/output are included in the characteristics of the input/output.
2. The distributed storage system according to claim 1 , wherein
types of the input/output, that is, random input/output and sequential input/output are included in the characteristics of the input/output.
3. The distributed storage system according to claim 1 , wherein
each node has the storage medium which stores the data, and
migration of the data between/among the nodes includes migration of the data between/among the storage media.
4. The distributed storage system according to claim 1 , wherein
the processor, the memory, the network and/or the storage medium are included in the resources whose usages are estimated.
5. The distributed storage system according to claim 1 , wherein
the processor
in a case where migration of the data is to be decided, decides an execution timing of migration of the data on the basis of the estimate of the per-period usage of each resource on the node which is taken for migration of the data.
6. A data migration method in a distributed storage system which includes a plurality of nodes which are connected with one another over a network, have the processors and memories individually and thereby provide a volume into/from which a host system inputs/outputs data and a storage medium which stores the data which is input into/output from the volume, comprising the steps of:
acquiring a per-period usage of each of a plurality of kinds of resources per data by the processor of any one of the nodes;
estimating a per-period usage of each resource on the migration destination node when the data is migrated to the migration destination node by using the acquired per-data resource usage and calculating an estimate of a latency from the estimated usage of each resource by the processor of any one of the nodes; and
deciding a migration pattern of the data on the basis of the estimate of the latency by the processor of any one of the nodes;
wherein the processor acquires the input/output together with characteristics of the input/output, and calculates a usage of each resource by the input/output on the basis of the characteristics of the input/output;
wherein types of the input/output, that is, read input/output and write input/output are included in the characteristics of the input/output.
7. The data migration method according to claim 6 , wherein
types of the input/output, that is, random input/output and sequential input/output are included in the characteristics of the input/output.
8. The data migration method according to claim 6 , wherein
each node has the storage medium which stores the data, and
migration of the data between/among the nodes includes migration of the data between/among the storage media.
9. The data migration method according to claim 6 , wherein
the processor, the memory, the network and/or the storage medium are included in the resources whose usages are estimated.
10. The data migration method according to claim 6 , wherein
the processor
in a case where migration of the data is to be decided, decides an execution timing of migration of the data on the basis of the estimate of the per-period usage of each resource on the node which is taken for migration of the data.