A Study Of Request Scheduling Optimization For QoS Guarantee In Multi-Tenant Clouds

Cloud computing is an important innovation resulting from the highly developed modern computer network technology and high-performance processor technology.It has become a strategic new industry.Due to the cost reduction and increased flexibility in resource allocation,more and more tenants are migrating their applications to the cloud.These tenants share the physical resources in the cloud computing platform,making the cloud computing platform exhibit multi-tenant characteristics.How to schedule requests and ensure the quality of service(QoS)in a multi-tenant cloud has become a concern for both cloud service providers and tenants.When users migrate their business to the cloud,cloud service providers need to schedule application requests and assign corresponding servers to process them.Then,service function chain(SFC)request scheduling is performed to ensure that traffic passes through network functions(NFs)in order for security and performance requirements.At last,SFC request updates are made to address network function and link overload caused by dynamic traffic.However,the current multi-tenant cloud exhibits three major characteristics:high application request volume,low NF reliability,and high traffic dynamism,which pose challenges to application request scheduling,SFC request scheduling,and SFC request updates.The high application request volume demands a fast response and high tenant isolation for application request scheduling,while the low NF reliability requires consideration of the impact of NF failures on SFC service reliability during SFC request scheduling.The high traffic dynamism necessitates that cloud service providers ensure that the configuration remains effective after SFC request updates.To address these challenges,this paper conducts research on request scheduling optimization for QoS Guarantee in multi-tenant clouds.The main research contents and contributions of this paper are as follows:1.With the increasing adoption of cloud computing by enterprises,the number of application requests submitted by users is also increasing.Existing fine-grained application request scheduling can achieve high resource utilization and maximize profits,but it has long scheduling delays and low scalability,making it unsuitable for large-scale application request scenarios.Moreover,it does not limit the number of tenants served by a server,resulting in poor tenant isolation.This paper designs a tenant-grained application request scheduling method,which aggregates requests belonging to the same tenant before scheduling them.By reducing the number of requests that need to be scheduled,the method significantly reduces service response time,improves system scalability,and ensures tenant isolation by allocating servers to tenants as much as possible.The paper formalizes the tenant-grained request scheduling problem as an integer linear programming problem and proposes approximate algorithms with a guaranteed approximation ratio for offline and online scenarios.Simulation and experimental results show that compared with fine-grained request scheduling methods,the proposed method can reduce scheduling overhead by about 90%while achieving similar throughput performance and ensuring tenant isolation.2.After application request scheduling,the servers that handle application requests are determined.To ensure network performance and security,the traffic generated by these servers needs to be processed by several network functions in a sequence to meet the requirements of the service function chain(SFC).The scheduling of traffic to the corresponding network functions for processing to meet SFC requirements is called SFC request scheduling.Due to potential hardware failures and software errors,network functions may fail.In multi-tenant clouds,a network function instance often serves multiple tenants,leading to uncontrolled tenant scope when the network function instance fails.This paper designs a high-robustness SFC request scheduling method that limits the number of tenants served by each network function instance during SFC request scheduling to limit the impact of network function failure on tenants.We also designed a traffic processing recovery mechanism that is independent of the control plane in the data plane to reduce recovery latency.Experimental results show that this approach can reduce the number of affected tenants by about 60%and lower the recovery latency by about 73%.3.Due to the dynamic nature of traffic,some links and network function instances in the network may become overloaded,necessitating the updating of SFC requests by selecting appropriate network functions and updating routing configurations.Given the high dynamism of traffic,the new routing configurations should be deployed within a limited time frame.Furthermore,considering the low reliability of NFs and the impact of malicious tenants on system robustness,it is necessary to ensure that the updated SFC routing configuration has robustness during the SFC request updating.To this end,this paper proposes a robustness-aware real-time SFC request update method,which enhances system robustness by limiting the number of network function instances that each tenant’s traffic can access and the range of tenants that each network function instance can serve during routing configuration updating,and ensures that the target routing configuration is deployed in real time to ensure the effectiveness of the routing configuration.Experimental and simulation results show that this approach can reduce update latency by about 70%and limit the scope of network function failure and malicious tenant impact.