TinyChecker: Transparent protection of VMs against hypervisor failures with nested virtualization

The increasing amount of resources in a single machine constantly increases the level of server consolidation for virtualization. However, along with the improvement of server efficiency, the dependability of the virtualization layer is not being progressed towards the right direction; instead, the hypervisor level is more vulnerable to diverse failures due to the increasing complexity and scale of the hypervisor layer. This makes tens to hundreds of production VMs in a machine easily risk a single point of failure. This paper tries to mitigate this problem by proposing a technique called TinyChecker, which uses a tiny nested hypervisor to transparently protect guest VMs against failures in the hypervisor layer. TinyChecker is a very small software layer designated for transparent failure detection and recovery, whose reliability can be guaranteed by its small size and possible further formal verification. TinyChecker records all the communication context between VM and hypervisor, protects the critical VM data, detects and recovers the hypervisors among failures. TinyChecker is currently still in an early stage, we report our design consideration and initial evaluation results.