46results about How to "Reduce the number of IOs" patented technology

The invention discloses a system and method for fast detaching a virtualized disk image. The system comprises a partition table analysis module, a disk analysis module, a file distribution table extraction module, a partition analysis module, a conflict resolution module and a file distribution table write-in module. The partition table analysis module is used for analyzing a target partition and a virtualized disk file physical address. The disk analysis module is used for judging whether a virtualized disk file is within the range of the target partition or not based on the physical address, if not, an error is reported, quitting is carried out, and if yes, the virtualized disk file is analyzed, and a file partition table and a file distribution table are acquired; the file distribution table extraction file is used for judging whether a file in the file partition table is stored in a virtualized disk as it is or not based on the file distribution table and recording file distribution records of files capable of being directly extracted; the partition analysis module is used for analyzing the target partition to acquire the position of a target partition file distribution table; the conflict resolution module is used for comparing the file distribution records with file names in the target partition, and if the file names are identical, file distribution record file names are modified; the file distribution table write-in module is used for writing the file distribution records in the target partition based on the target partition file distribution table and deleting original files.

The embodiment of the invention provides a big data write-in method and device, a big data read method and device and a big data processing system. The write-in method includes the steps that a write-in request is received; big data are segmented into multiple pieces of subdata according to the write-in request; the subdata are uniform in length, and relative positions of data parts, corresponding to the same field, in the subdata are fixed; the subdata are written in a specified storage medium; storage positions, in the storage medium, of current subdata are added to all the subdata. The big data can be positioned easily, conveniently and rapidly, time consumption is reduced, the speed of storing and reading the big data is increased, and system performance and working efficiency are improved.

The invention discloses a large-scale social network-oriented graph data storage and query method. The data storage manager of the invention stores the received graph data in a Key-Value manner, uses the vertex ID of the graph data as the Key, and uses the vertex neighbor The field is Value; data storage for each vertex neighborhood: store multiple edges connected to the vertex neighborhood in a fixed-size memory block in order with time stamps, and form a doubly linked list, and attribute information of the vertex and index information are stored in a data structure. When the data storage manager receives an access request to visit vertex v, the data storage manager transmits the vertex v and its k-order neighbors to the requester; the requester caches the returned data locally, and the next query, first check The local cache, if there is no queried vertex, sends the access request to the data storage manager. The invention can satisfy dynamic update, and is suitable for processing sparse data scenarios and random access.