Gating Mechanism For Real-time Network I/O Requests Based On Para-virtualization Virtio Framework

Cloud computing is a computing service.Compared with the traditional IT service mode,cloud computing can guarantee the immediacy of user needs based on the Pay-as-you-use.Among them,service level agreements(SLA),which are used to standardize the services of cloud computing service providers and protect users' rights and interests,have been paid more and more attention.Simply,SLA is a contract or agreement with multiple requirements regarding service quality established through negotiation between the service provider and the user,including the Service Level Objective(SLO).In the cloud computing service environment,users read and write data,and the generated network I/O requests(hereinafter referred to as “requests”)access the underlying resources through the Virtio framework.Due to intensive user requests and limited system resources,Response Time(RT)is an important performance indicator for measuring service quality in SLO.It is related to the amount of resource usage.If more resources are allocated to the request,the requested performance and short response time will be guaranteed.If the resources allocated to the request are insufficient,the request will wait for resources,and the response time will increase.Therefore,how to allocate resources for cloud computing services to ensure the response time target specified by SLO has become a key technology.The commonly used method is to ensure response time by controlling resource usage.The problem with this method is that when resources are insufficient,the response time of the request cannot be guaranteed,and different requests to access resources are difficult to be treated differently.For this reason,a method to ensure the response time of the request by controlling the access of the request to the resource is proposed.This method establishes a gating mechanism inside or outside the system to control the time and frequency of the request to access the resource or enter the system.However,most of the gating mechanisms within the system need to modify the system,and it is also necessary to select appropriate resources as the gating objects that request access.However,the gating mechanism outside the system is not controlled after the request enters the system,and it is difficult to guarantee the request response time.Therefore,it is the key to find a location that has little impact on the system and is sensitive to response time to establish a gating mechanism.This paper adopts the method of controlling the access of the request to the resource to ensure the response time of the request.Based on the paravirtualization framework Virtio of the KVM platform,this paper proposes the real-time network I/O request gating mechanism(GMRNR).Through the establishment of a gating mechanism to control the access of network requests to data,ensuring the response time of the request.The innovations of this paper are as follows:(1)GMRNR is established in the Virtio front-end virtio-net module to intercept network I/O requests,control direct transmission of high-priority requests,aggregate transmission of low-priority requests,and prioritize the response time of high-priority requests to the greatest possible extent Strive to ensure response time for low-priority requests.The establishment of a gating mechanism in Virtio has little impact on the system,is convenient to modify the drive,and is direct and effective in system control.It can control the time and frequency of different priority requests to access data,thereby ensuring the response time of the request.(2)GMRNR,using two-end aggregation method(TAM),classifies and aggregates the different quests according to priority after taking into consideration the importance of the request and response time requirements.Requests with high importance and high response time requirements have high priority.Conversely,requests with low importance and low response time requirements have low priority.Through classification and aggregation,requests are classified and processed to share access to data,reducing system switching and reducing CPU overhead.(3)GMRNR uses timers and aggregation queue length to control the time when requests of different priority levels pass through Virtio and the frequency of aggregation to ensure the response time of requests of different priority levels.After different priority requests is set different timers,the latest time of aggregation transmission will be presented according to the required response time.At the same time,after aggregation queues of different lengths are set,high priority requests will accompany a short aggregation queues and high aggregation frequency.Conversely,requests with low priority have a long aggregation queue and low aggregation frequency.When the length of the aggregation queue reaches the threshold or the request in the queue reaches the latest transmission time set by the timer,the request is transmitted to the backend for processing.Experiment tests show that: GMRNR can distinguish the priority of network I/O requests.When resources are sufficient,different levels of network I/O requests can be completed within their respective required time;when resources are insufficient,priority is given to ensuring high priority network I /O request response time.At the same time,GMRNR has higher resource utilization efficiency.