A media company runs a nightly batch video-rendering job that can be paused and resumed at any time. The workload is fault-tolerant and the jobs can be restarted if interrupted. The company wants the lowest possible EC2 compute cost. Which EC2 pricing model best fits this requirement?