Method for multiple virtual machines to access distributed object storage system

Abstract

The invention relates to a method for accessing a distributed object storage system through a mode of designing a local cache in a client under the architecture of the distributed object storage system. The method comprises a reading operation: when reading the object file data of the distributed object storage system, firstly, searching an object file in the local cache of the client, if the object file is not searched in the local cache, reading the object file from the distributed object storage system, storing in the local cache system of the client, if the object file is searched in the cache, not reading the object file from the distributed object storage system; writing operation: when writing object field data in the distributed object storage system, firstly writing the object file data in a local cache, after finishing writing, marking that the object file is not synchronized, returning, and then writing the object file data written in the local cache in the distributed object storage system by a cache synchronous service module. Through the invention, the number of access times to the distributed object storage system can be reduced; the I / O throughput of accessing the file system by the virtual machine is improved; and the response time of an access request is reduced.

Description

technical field [0001] The invention belongs to the technical field of computer storage, and specifically relates to constructing a local cache mechanism in each virtual machine, reducing the times of accessing a distributed object storage system, reducing the access bandwidth of a storage network, and improving access efficiency. Background technique [0002] The data access of multiple virtual machines in the cloud computing environment requires the Distributed Object Storage System (Distributed Object Storage System: DOSS) to provide high throughput and fast response time. Distributed object storage systems have made great progress. Distributed object storage systems, represented by Ceph, Sheepdog, GlusterFS, Luster, etc., use multiple network nodes to provide distributed storage capabilities and realize massive data storage. Storage performance can grow nearly linearly with the increase of storage nodes. However, the current distributed object storage system has the fo...

AI Technical Summary

Problems solved by technology

[0003] (1) Although its total bandwidth will increase with the increase of storage nodes, for a single DOSS client, the total bandwidth is subject to the client's network bandwidth

[0004] (2) The I / O data of the distributed object storage system is transmitted through the network, which makes the bandwidth load of the transmission network too high

If multiple client software accesses DOSS, it cannot get enough network bandwidth and cannot support enough IOPS (the number of read and write (I / O) operations per second)

[0005] (3) The distributed object storage system DOSS retains multiple copies of data. When writing / updating operations, multiple copies need to be written / updated to ensure the storage consistency and data reliability of DOSS, which reduces the writing performance of DOSS.

Building a cache area inside the distributed object storage system DOSS can improve the read and write performance of DOSS, but it cannot reduce the communication volume between the DOSS client software and the distributed object storage system DOSS, which will consume a lot of network bandwidth; Moreover, it is necessary to add a cache device on each storage node, which increases the construction cost of the distributed object storage system, and has limited acceleration effect on hot data reading and writing.

Images