Indy Gray Sort and Indy Minute Sort

Indy and Daytona are two very different kinds of benchmark on the large cluster. Former system tends to treat Indy as a special case of Daytona and using the same code and configuration with little modification. However, the less restricted Indy sort is actually an attempt to archive the theoretical upper limit of the hardware, whatever the bottleneck might be. While Daytona need a system providing fault tolerance which will degrade the performance inevitably. We would rather explore the hardware limit as the target cluster contains thousands of machines with large memory, so we aim at Indy instead of Daytona. We find TritonSort[1] was a welldesigned system with main focus on disk performance and we extend their experiments from 52 machines to thousands.