4 We Are Looking To Run Three Computations Or Jobs On The Iridis 5 Supercomputer At The University Of Southampton J1 1 (25.46 KiB) Viewed 121 times
4 We Are Looking To Run Three Computations Or Jobs On The Iridis 5 Supercomputer At The University Of Southampton J1 2 (39.03 KiB) Viewed 121 times
4. We are looking to run three computations (or jobs) on the IRIDIS 5 supercomputer at the University of Southampton: j1 j2, j3. A total of 5 compute nodes are available: n1, 12, 13, 14, n5. Jobs j1 and j2 carry out some complex matrix computations and can only be run on compute nodes n₁ and n2 (which are the only ones equipped with a GPU). Job ją can run on any compute node. The jobs can be seamlessly moved
from a compute node to another one as many times as necessary (for instance, a job could run for 20% of the time on n₁, for 40% of the time on ne, and for 40% of the time on n3). Since IRIDIS 5 is oversubscribed, each compute node can be used for no more than 7 days in total and a single job can be scheduled to run on a node for, in total, no more than 2 days. For their completion, 31, 32, and j3 need, respectively, 4, 5, and 6 days of computations. Model the problem of scheduling the three jobs on the IRIDIS 5 supercomputer as a maximum-flow problem and report the corresponding graph (with source, sink, and arc capacities). [20 marks]
Join a community of subject matter experts. Register for FREE to view solutions, replies, and use search function. Request answer by replying!