Cost-efficient live VM migration based on varying electricity cost in optical cloud networks

As cloud services have increased in popularity, Information and Communication Technology (ICT) has been assuming a larger share of worldwide energy consumption. Energy cost resulting from ICT energy consumption contributes a significant part of the operating expenditure for cloud-service providers. We note that the energy cost at a geographical location has a direct correlation to the price of electricity (which may vary with time depending on the location). Different geographical regions of a country, e.g., the USA, may have different electricity prices with variation in prices from one time of the day to another. We show that this spatial and temporal variation in electricity price can be exploited to reduce the operating cost associated with energy consumption. Compute workloads can be relocated to regions with cheaper electricity prices. We consider a multi-datacenter (DC) environment where workloads are virtualized into virtual machines (VMs). In order to move VMs from one DC to another, we employ the live VM migration approach. Our model for cost-efficient VM migration is based on varying electricity prices such that it migrates VMs to DCs with cheaper electricity prices while simultaneously considering multiple parameters such as bandwidth for migration, cost of migration, duration of migration, and number of servers and racks to be switched on/off during each migration period. We find that our model significantly reduces the energy cost associated with cloud-service operation.