Towards a SLA-compliant Cloud Resource Allocator for N-tier Applications
Aaron McConnell, Gerard P. Parr, Sally I. McClean, Philip Morrow, Bryan Scotney · 2012
Abstract—Cloud vendors commonly offer users IaaS where vir-tual machines (VMs) can be created and run on cloud resources. The resource allocation for each VM is defined by the user and the VM is created on a physical machine (PM) where ample resources exist to support the VM’s operation at its maximum capacity. There are a number of opportunities for improvement when allocating host resources to VMs. VM-resident applications are often n-tier, with different VMs responsible for parts of the distributed application. It may be important that these VMs are placed within a given network proximity to one another. The network proximity to the user may also be an issue for some applications. Resource allocation to VMs should also be such that, rather than a user over-provisioning the VM, the VM’s minimal operational requirements are specified so that the VM can be resource-throttled at times of heavy load. This paper presents an outline for a system called Innkeeper, which aims to allocate resources to a VM in a way that ensures the VM will always function adequately, but where the VM is not over-provisioned. Innkeeper also aims to place VMs so that a VM ”family ” are kept within a necessary network proximity to one another and where the proximity to the user is also considered when placing VMs. Keywords-cloud computing; resource allocation; virtualisation. I.