hyperCache: A Hypervisor-Level Virtualized I/O Cache on KVM/QEMU

While VDI offers several benefits, such as the increased resource utilization and the private data protection, there exist problems that can deteriorate the system performance, including I/O virtualization overhead. Before I/O requests issued in VMs are completed in VDI, they should go through multiple software layers, such as the layer from the backend, shared storage to the host server and the layers between guest operating system, hypervisor, and eventually host operating system. In this paper, we present a hypervisor-level cache, called hyperCache, which is possible to provide a shortcut in KVM/QEMU by intercepting I/O requests in the hypervisor, while taking into account the I/O access frequency. Our experimental results demonstrate that our design improves I/O bandwidth over the existing QEMU.