Adaptive Quality of Service (QoS) using OpenFlow

Abstract

Server Virtualization and Cloud based data centre solutions are becoming popular for increasing resource utilization, better manageability and cost reduction. Server virtualization allows us to run multiple independent instances of operating systems i.e. virtual machines (VMs) on a single physical server. Physical resources of server like CPU, memory, storage and network is shared among multiple VMs, and thus it maximizes the resource utilization. But on the other side, advantage of resource utilization may also raise performance issues if these resources are not managed and provisioned in a proper way among VMs. Network resource is crucial for the performance of network based applications. Sometime despite of having enough CPU and memory resource, application service delivers degraded performance to end user just because of in-su cient provisioned network resource to the VM. Therefore e cient network resource allocation technique is required for application QoS. Network resource can be provisioned proactively to VMs, but it requires advance knowledge of application user load pattern based on statistical analysis. This technique does not work e ciently because application user load is highly dynamic and does not always follow the predictable pattern. We propose an adaptive QoS mechanism using reactive approach, which monitors the application user load in a real time and provision network resources accordingly. More network resources are provisioned to most demanding applications and vice versa. OpenFlow based Software De ned Networking (SDN) technology is used to meet the objective. Our test result shows that Quality of Experience (QoE) targets of maximum user sessions are met which belongs to demanding application.