A storage system and method and device for placing IO disks therefor

Abstract

The invention discloses a method, device and storage system for IO disk dispatching of a storage system. The method includes: obtaining the maximum IO delivery rate according to the used cache ratio of the cache; according to the maximum IO delivery rate and preset fragmentation size, obtain the number of shards, and generate shards with the number of shards in the cache; store each received IO data on a corresponding shard; drop the shards that meet the preset placement conditions. The present invention simplifies the process of IO merging by forming a large IO data on each shard into a large IO for disk placement, reduces the hysteresis of the drop delay of IO data in batches, and meets the requirements of all-flash storage batches. On the basis of the IO data of disk placement, it is ensured that the upper limit of the actual rate is consistent with the upper limit of the rate of the leaky bucket current limiting algorithm, so that the subsequent disk placement delay of shards can be used to dynamically adjust the rate and accurately control the IO disk placement.

Description

technical field [0001] The present invention relates to the technical field of storage systems, and in particular, to a method, device and storage system for dropping IO disks in a storage system. Background technique [0002] With the development of science and technology in modern society, the application of storage systems is becoming more and more extensive. At present, the dynamically balanced leaky bucket current limiting algorithm commonly used in traditional storage systems performs IO (Input / Output input / output, that is, read / write) off-disk, that is, sends IO data to disk. The core idea is as follows: According to The proportion of the used cache determines the maximum rate of IO data delivery, and the response time of the lower layer is counted, and the rate is adjusted on the basis of the maximum rate to achieve a balance between data request and processing speed. However, in the emerging all-flash products, the log-type disk placement method is adopted, so that...

AI Technical Summary

Problems solved by technology

However, in the emerging all-flash products, a log-type storage method is adopted, so that multiple small IOs must be merged into a large IO through a complex process before they can be placed, that is, batch IO data; due to the time introduced by IO merging delay, so that the existing statistics of small IO response delay indicators cannot truly reflect the delay of data placement; and due to the complexity of the existing IO merging process, the scheme based on large IO statistics of placement time has a certain lag. As a result, the actual upper limit of the rate is inconsistent with the rate limit of the leaky bucket current limiting algorithm, especially in scenarios where the proportion of the used cache varies greatly, and it is difficult to accurately control IO placement

Images