A VM based hosting architecture system in which finer grain control in optimizing multiple workloads across multiple servers is provided. The system includes a plurality of servers to be utilized by multiple workloads. In addition, the system includes a plurality of virtual machines (VMs) at each of the plurality of servers, wherein the plurality of VMs at each of the plurality of servers each serve a different one of the multiple workloads. Moreover, the system includes resource management logic to distribute server resources to each of the plurality of VMs according to predicted resource needs of each of the multiple workloads. Each of the multiple workloads are distributed across the plurality of servers, wherein fractions of each of the multiple workloads are handled by the plurality of VMs. The distribution of multiple workloads over multiple servers has the effect of achieving a finer grain control in optimizing workloads across the plurality of servers.