You are managing a nightly batch workload that relies on a significant number of virtual machines (VMs) to complete its tasks. This workload is designed to be fault-tolerant, meaning it can handle scenarios where a few VMs are unexpectedly terminated without compromising overall performance. However, the current expense associated with running these VMs is proving to be exorbitant. What strategies can you implement to reduce the cost of these VMs while maintaining the fault-tolerant capability of the workload?