Power Aware Load Balancing for Cloud Computing
Jeffrey Galloway, Karl Smith, Susan S. Vrbsky · 2011
With the increased use of local cloud computing architectures, organizations are becoming aware of wasted power consumed by unutilized resources. In this paper, we present a load balancing approach to IaaS cloud architectures that is power aware. Since the cloud architecture implemented by local organizations tends to be heterogeneous, we take this into account in our design. Our Power Aware Load Balancing algorithm, PALB, maintains the state of all compute nodes, and based on utilization percentages, decides the number of compute nodes that should be operating. We show that our solution provides adequate availability to compute node resources while decreasing the overall power consumed by the local cloud by 70% - 97% compared to using load balancing techniques that are not power aware. Cloud computing architectures are becoming a dominant contender in the distributed systems paradigm. Using this architecture, customers are given access to resources provided by a cloud vendor as described in their Service Level Agreement (SLA). Clouds use virtualization technology in distributed data centers to allocate resources to customers as they need them. Generally, clouds are deployed to customers giving them three levels of access: Software-as-a-Service (SaaS), Platform-as-a-Service (PaaS), and Infrastructure-as-a-Service (IaaS). The jobs can differ greatly from customer to customer. Each commercial vendor has a targeted customer and specific markets they wish to saturate. Local cloud implementations are becoming popular due to the fact that many organizations are reluctant to move their data to a commercialized cloud vendor. There are debates on whether moving data to the public cloud would benefit small organizations. Beyond the question of benefit to the organizations utilizing public clouds, there are also issues with trust, security and legality. Some organizations may not trust a third party with their information and/or software. Other organizations may not be comfortable allowing a third party to be responsible for the security of the cloud.