Аннотация:HPC-as-a-service is a new cloud paradigm that represents easy to access and use cloud environments for High Performance Computing (HPC). This paradigm has been receiving a lot of attention from the research community lately since it represents a good tradeoff between computational power and usability.One of the key drawbacks associated with HPC clouds is low CPU usage due to the network communication overhead. Instances of HPC applications may reside on different physical machines separated by significant network latencies. Network communications between such instances may consume significant time and thus result in CPU stalls.In this paper we propose the scheduling algorithm that overcomes such drawbacks to increase the HPC task capacity in the Ethernet-based HPC cloud by sharing CPU cores between different VMs. The algorithm observes parallel tasks’ behavior and packs tasks with low CPU usage on same CPU cores.We fully implemented and evaluated our algorithm on 15 popular MPI benchmarks/libraries. The experiments have shown that we can significantly improve the CPU usage with negligible performance degradation.