A distributed block storage implementation method for a virtual machine

Abstract

The invention discloses a method for implementing distributed block storage of a virtual machine, which is applied to a distributed storage system including a master control node, several client nodes and data storage nodes. The client nodes map virtual disks to virtual machines and The master control node sends a request to operate the virtual disk, the master control node processes and stores the information of the virtual disk, the data storage node provides physical storage space for the virtual disk, and the implementation method of the distributed block storage of the virtual machine includes: the master control node receives and creates the virtual disk The command selects the data storage node to create a fragment based on preset rules, and sends the corresponding data storage node address to the client node; the client node creates a fragment on the data storage node based on the received address, and the fragment file is in Qcow2 format. A virtual machine distributed block storage implementation method proposed by the present invention virtualizes multiple Qcow2 files into a large virtual disk and can simultaneously satisfy I / O balance without hotspots, has high performance, and is simple and easy to maintain.

Description

technical field [0001] The invention relates to the field of distributed storage of virtual machines, in particular to a method for realizing distributed block storage of virtual machines. Background technique [0002] With the rapid development of cloud computing technology, Infrastructure as a Service (IaaS), as the base of cloud computing, is becoming more and more important. Virtual machine service is the core of IaaS, so the status and requirements of providing storage services for virtual machines are getting higher and higher. [0003] Virtual machines of cloud operators need storage that supports high reliability, scalability, scalability, and low cost. Traditional virtual machine storage services can be divided into three categories: direct-attached storage (DAS) for open systems, network-attached storage (NAS) and storage area network (SAN). However, traditional storage cannot meet the storage requirements of virtual machines in IaaS scenarios. First of all, it ...

AI Technical Summary

Problems solved by technology

[0007] 1) Its data distribution cannot be fully controlled. For example, in order to improve reliability, data is usually stored in three copies, one of which is required to be stored in SSD (in order to improve read performance), and the other two are stored in mechanical hard disk. Then sheepdog is impossible

2) When there are fewer storage nodes, the hash algorithm is prone to data imbalance (the amount of data stored by each node varies greatly)

3) When a node goes down, it will cause the load of adjacent nodes to increase

[0008] 2. A large number of fragments lead to insufficient storage performance

When opening, the kernel needs to find the storage location of the fragment in the file system, and when closing, the operating system needs to refresh the cache of the fragment (file) to the physical disk, so the performance of this method is low

[0009] 3. A large number of fragments lead to complex management

When some hosts are abnormal, if the software cannot automatically recover, it is almost impossible to achieve manual recovery in the face of massive fragments

Images