Multi-virtual-machine mapping and multipath fuse acceleration method and system based on kvm

Abstract

The invention provides a multi-virtual-machine mapping and multipath fuse acceleration method and system based on a kvm. The method comprises the steps that 1, a delayed writing function is added for a data path of an existing FUSE kernel module, wherein the delayed writing function is that data is aggregated in the FUSE kernel module through a VFS layer and is directly returned to the system for calling when passing through a kernel, and the data is transmitted to a user-mode client-side for data trading operation only when the aggregated data meets a certain size or is added no longer in a certain timer period; 2, multi-mounting-point mounting is conducted on the FUSE kernel module. Multi-virtual-machine data is mapped to a host machine by stripping the FUSE module function and a multi-virtual-machine mapping mechanism, the handling operation on a stripping client-side is issued to a host machine side to reduce the pressure of virtual machines, and more resources are provided for the virtual machines to perform calculating and processing tasks.

Description

technical field [0001] The invention relates to the technical field of file systems, in particular to a kvm-based multi-virtual machine mapping and multi-path fuse acceleration method and system. Background technique [0002] At present, there are mainly two ways to use the file system client: one is the API library, like the HDFS file system, which is written in JAVA and provides API and shell scripts for users to use. The existing programs of the user cannot be used directly, and the API function of HDFS is required. can be used, and the scope of use is limited. The first is POSIX semantics, that is, the file system is mounted in the local directory by mount, just like operating local files. Compared with API and SHELL methods, the file system that supports POSIX semantics greatly facilitates the use of users. According to the client encoding method, It is further divided into kernel state file system client and user state file system client. The kernel state file system ...

AI Technical Summary

Problems solved by technology

[0005] Compared with the kernel-mode file system, the user-mode file system based on FUSE has the following disadvantages: Through testing, it is found that the user-mode file system has lower IOPS than the kernel-mode file system, and the read and write speed of the user-mode file system is extremely slow. , which greatly affects the performance of the system; FUSE is divided into two parts, the kernel part and the user-mode lib library. The kernel FUSE module is extremely dependent on the system kernel function, which is incompatible with each other. It must be updated to use the kernel, which makes the use of FUSE very uncomfortable. convenient

During the small-grained sequential write test of 4kB data, the IO of the system was only 79MB / s, and during the sequential write test of 1MB data with large granularity, the system IO bandwidth was only 161MB / s, which greatly limited the system performance.

Images