Virtual I/O scheduling method and system

Abstract

The invention discloses a virtual I / O scheduling method and system. The method includes the steps that when an upper-level application generates an I / O request Rk for a file system, a CFD captures the request Rk through a filter driver of the file system, classifies the request according to a file Fk, a process Pk and a timestamp Tk, adds a class label Ck into the request Rk and inserts the I / O request into an I / O queue of the corresponding class according to the added class label Ck; data blocks needed by the I / O request are searched for in a local cache, and if the data blocks are found, the process returns. By means of the virtual I / O scheduling method and system, low overhead and little modification on an operating system are achieved, and the problem that in a cloud calculation environment, numerous confusing types of I / Os generated by terminal equipment lead to poor system performance is solved.

Description

technical field [0001] The invention relates to the field of cloud computing, in particular to a virtual I / O scheduling method and system. Background technique [0002] Today's computing era is transitioning from information sharing to service sharing, and a new computing model - cloud computing - has emerged in this leap. Cloud computing supports users to find relevant computing services from various devices (including fixed, mobile and household devices) according to their own needs, and these services are stored in servers located in distributed networks. Cloud computing is a networked and on-demand distributed system, and the terminal system will generate a large number of mixed types of virtual I / O. It includes system I / O requests, non-real-time file I / O requests and real-time file I / O requests, etc. figure 1 Shown is the processing process of virtual I / O in the cloud computing environment. However, the rapid improvement of CPU computing speed in recent years has not...

AI Technical Summary

Problems solved by technology

[0005] Second, the I / O scheduling algorithm that supports QoS control requires a QoS control module that supports hard real-time in the operating system to provide QoS control parameters, which will bring a large overhead from the application program down to the physical disk scheduling, and for The modification of the operating system is also relatively large, and the implementation is complicated

[0006] Third, cloud computing adds network transmission and I / O processing on the server to the path of processing virtual disk I / O. The path is longer than that of a stand-alone system, so small data packets generated by a large number of small I / O will make network transmission The performance of network I / O is reduced, and the performance of network I / O is reduced

Existing methods for aggregating small I / O, such as Collective I / O, cannot obtain the semantic information of the upper layer, that is, the type of I / O is not distinguished, so the efficiency of aggregation is not high

Images