Cypress: A Scheduling Infrastructure for a Many-Core Hypervisor

In this position paper, we present our vision for the scheduling infrastructure in a many-core hypervisor – the hypervisor targeted for many-core platforms. The key objectives of our system are scalability and heterogeneity-awareness. We see these as firstorder objectives, because future many-core processors will consist of thousands of cores and those cores will be heterogeneous. Since existing hypervisors were not designed to handle the scale and heterogeneity of many-core hardware, our design will differ from that of existing hypervisors in many important ways. The design of our experimental many-core hypervisor, Cypress, is based on three principles: partitioning, localization, and customization. Together, these principles facilitate scalability, by minimizing the sharing of scheduling runqueues, and manage heterogeneity, by assigning to each VM the cores most suitable for its workload. In this paper we motivate our design, present its key components, discuss challenges in our research, and report on its status.