GiantVM: A Novel Distributed Hypervisor for Resource Aggregation with DSM-aware Optimizations

We present GiantVM, 1 an open-source distributed hypervisor that provides the many-to-one virtualization to aggregate resources from multiple physical machines. propose techniques enable CPU and I/O shared memory (DSM) achieve aggregation. GiantVM is implemented based on state-of-the-art type-II QEMU-KVM, it can currently host conventional OSes such as Linux. (1) identify performance bottleneck of be DSM, through a top-down analysis. Although offers great opportunities for CPU-intensive applications enjoy aggregated resources, memory-intensive could suffer cross-node page sharing, which requires frequent DSM involvement leads collapse. design guest-level thread scheduler, DaS (DSM-aware Scheduler), overcome bottleneck. When benchmarking with NAS Parallel Benchmarks, boost up 3.5×, compared default Linux kernel scheduler. (2) While evaluating DaS, we observe advantage resource reallocation facility. Thanks SSI abstraction migration done by scheduling. allows standby pages in destination, need not transferred network. The saved network bandwidth 68% average, VM live migration. Resource increases overall utilization 14.3% co-location experiment.