Enabling Ideal Job Mapping on Wireless Supercomputers and Datacenters

A number of small parallel applications run on datacenters and supercomputers simultaneously. Job mapping becomes crucial to improve system utilization and user experience especially on a supercomputer, because fragmentation of unused compute nodes could not be assigned for an incoming job with even a smaller size. Wireless supercomputers and datacenters with 60GHz radio or free-space optics have been proposed so that a diverse application workload can be better supported by changing network topologies by swapping the endpoints of wireless links. In this study we propose the use of wireless links for the purpose of giving an ideal job mapping. Our simulation results demonstrate that a large number of fragmentations heavily degrade the system utilization of supercomputers and datacenters under a conventional fixed network topology. By contrast, wireless interconnection networks can have an ideal job mapping by directly reconnecting non-neighboring computing nodes. It improves the system utilization by up to 17.7% for user jobs on a supercomputer. Thus the whole service time can be shortened especially for dealing with dozens of intensively incoming jobs. Furthermore, we confirmed that either workload or scheduling policy does not impact the fact that the ideal job mapping on wireless supercomputers outperforms that on wired 2-D mesh networks in terms of system utilization and whole service time.