Traffic aware workload provisioning in Cloud
Publication Date: 2017-Jun-01
The IP.com Prior Art Database
Traffic Aware Workload Provisioning in Cloud
In any existing cloud orchestration, virtual machines (VMs) are spawned based on a round-robin or another existing methodology. Spawning of the VM(s) (or containers) is not dependent on traffic situation on existing VMs on the system. The issue happens that the tenants are placed in a noisy neighbor situation. The main reason for not doing so is that the orchestration system does not care about the run time network load/situations. A runtime system is needed that can load balance the spawning of the VM(s) based on the traffic load at that time. The novel contribution is a protocol in which the cloud orchestration system talks to the Software-Defined Networking (SDN) controller before spawning the VMs on the hypervisor; therefore, it is aware of the load level on that particular hypervisor before it can spawn the VM. The method introduced herein enables any orchestration system to talk to the SDN system and receive a snapshot of traffic load per hypervisor, before the orchestration system can give a go-ahead for the VM on that hypervisor. When the orchestration engine receives the snapshot of traffic per hypervisor, it can eliminate some of the hypervisors that are having more than a given percentage (x%) of traffic usage. The steps for implementing the protocol and method in a preferred embodiment are:
1. SDN controller communicates with the existing hypervisors as per the existing protocol
2. SDN controller continually polls the...