Data center network ubiquitous caching method

Abstract

The invention provides a data center network ubiquitous caching method which includes the following steps of S1, sending a file reading request to a data center network through a user host computer, wherein the file reading request includes file names; S2, searching for original addresses of files, corresponding to the file names, in a plurality of host computers of the data center network according to the file names, and reading the files, corresponding to the original addresses according to the original addresses; S3, storing the read files in caches of the user host computers, and generating a cache control protocol; S4, when other user host computers request for reading the files, the files are obtained from the plurality of user host computers of the data center network according to data forwarding rules. According to the method, a cache host computer which is closest to a requesting user host computer is selected through the cache control protocol, and data are read from the cache host computer, occupation of a network flow is reduced, meanwhile, an integrated controller does not need to participate in the process, and therefore expandability of data centers is ensured.

Description

technical field [0001] The invention relates to the field of network technology, in particular to a data center network ubiquitous caching method. Background technique [0002] With the expansion of the data center, not only the number of hosts and switches has increased dramatically, but also the number and types of applications have also increased, which poses a more severe challenge to the performance of the data center network. Common content-based data center services include distributed computing engine MapReduce and distributed file system GFS, etc., which have a great demand for network bandwidth. Therefore, network bandwidth has become the main bottleneck affecting data center performance. [0003] Existing solutions mainly focus on increasing the number of links and link bandwidth. New switch-centric topology, such as Fat-Tree (fat tree), VL2 (virtual link layer) or host-centric topology, such as BCube (a high-performance data center topology), DCell (A recursive...

AI Technical Summary

Problems solved by technology

However, due to the high load of the centralized controller (controller) of the data center file system, the caching mechanism on the host needs to manage data storage, cache consistency, and cache data publishing by itself, and cannot add any additional burden to the controller.

Images