A time label based I/O scheduling QoS method

Abstract

A time label based I / O scheduling QoS method System Design Based on Time Label is provided. In the form of a two-level mapping queue, where the first level is the client queue for the client, the second level is a real request queue, each request contains three time stamps (Ri, Wi, Li), where i denotes the client number to which it belongs, and a large amount of dynamic data is organized and managed by constructing reserved time stamps, weighted time stamps and full binary trees of upper bound time stamps respectively. The I / O scheduling QoS method based on the time label designed by the invention can more reasonably allocate the limited I / O resources according to requirements, so as to cope with the flexible and changeable application scenarios and ensure the reliability of the system performance.

Description

technical field [0001] The invention belongs to the field of cloud storage of distributed big data, and in particular relates to a time tag-based I / O scheduling QoS method. Background technique [0002] With the gradual rise of cloud computing and virtualization technologies, the problem of I / O resource allocation in clusters has become increasingly serious. For example, in the production environment, some nodes may occupy most of the I / O resources of the entire cluster, resulting in long operation delays of other nodes and unfriendly user experience. QoS originated from network communication, which means that a communication network can use various basic technologies to provide better service capabilities for specific applications. For a system, QoS aims to coordinate the limited I / O resources of the system more reasonably, so as to realize resource allocation on demand and provide better external services. [0003] Traditional systems usually use token bucket-based prior...

AI Technical Summary

Problems solved by technology

[0003] Traditional systems usually use token bucket-based priority queues (prio) for I / O scheduling, but the tokens in the token bucket are a variable quantity, and the tokens in the token bucket need to be refreshed continuously, resulting in refreshing the same The time interval between the queue corresponding to the token in the token bucket is too long

In addition, if the weight-based priority queue (wpq) is used for I / O scheduling, the weight value of the algorithm is fixedly allocated, and dynamic bandwidth resource allocation cannot be realized according to the load status of the network.

The above-mentioned I / O scheduling algorithm has obvious shortcomings, and cannot really play the characteristics of QoS

Images