This paper analysis various challenges faced in optimizing computing resource utilization via load balancing and presents a platform-independent model for load balancing which targets high availability of resources, low SLA (Service Level agreement) violations and saves power. To achieve this, incoming requests are monitored for sudden burst, a prediction model is employed to maintain high availability and a power-aware algorithm is applied for choosing a suitable physical node for a virtual host. The proposed dynamic load balancing model provides a way to conflicting goals of saving power and maintaining high resource availability.For anyone building a private, public or hybrid IaaS cloud infrastructure, load balancing of virtual hosts on a limited number of physical nodes, becomes a crucial aspect. This paper analysis various challenges faced in optimizing computing resource utilization via load balancing and presents a platform independent model for load balancing which targets high availability of resources, low SLA (Service Level agreement) violations and saves power. To achieve this, incoming requests are monitored for sudden burst, prediction model is employed to maintain high availability and power aware algorithm is applied for choosing a suitable physical node for virtual host. The proposed dynamic load balancing model provides a way to conflicting goals of saving power and maintaining high resource availability.