File storage dynamic aggregation based optimization method

Abstract

The invention discloses a file storage dynamic aggregation based optimization method. In the optimization method, a copy of a file which is frequently accessed by a node is acquired and is packed and stored at a local node by dynamically combining the file, so that frequency of remotely reading the file by the node is reduced and user data access delay is reduced. The method comprises the following steps of: 1, determining whether data is overlapped on the basis of a data original, detecting whether a request file sent by a user is in a local node, if the data is overlapped, judging that the request file is in the local node, if the data is not overlapped, turning into the step 2; 2, detecting whether sufficient storage space is locally available to accommodate a request file based on a copy placement optimization algorithm, if yes, selecting a remote node with minimum performance cost to obtain a request file copy, is not, turning into the step 3, wherein the cost comprises file transmission time and queuing time on the remote node; and 3, based on a storage space optimization algorithm, calculating times t1 of reading the request file by the node, calculating times t2 of readingless files from all files stored in the local node, if t1 is more than t2, deleting the t2 and clearing space for storing the request file. The cost of the copy selected to be replicated by the method is the minimum at the present and the cost is also the minimum in later time.

Description

technical field [0001] The invention belongs to the field of distributed storage, in particular to an optimization method based on file storage dynamic aggregation. Background technique [0002] Distributed storage has established an integrated storage, processing, access, transmission and service architecture and heterogeneous distribution environment for massive data. Data replication can not only improve the efficiency of data access, but also improve the load balance and reliability of the system. [0003] The traditional replication method is to replicate a data master to replicas at different sites. The reason for this situation is that some past or present replication requirements and processing procedures are relatively simple. With the development of data replication technology, the situation has undergone many changes. Different replication source copies can be divided on the same data source object, and these different replication source copies have their own mult...

AI Technical Summary

Problems solved by technology

There is such a problem in this model when evaluating the value of data replication: nodes often make decisions according to their own interests, so they may not necessarily get the best global benefits

Images