CAREER: Hardware Error Resilient Virtualization Infrastructure

George Washington University, Washington DC

Investigators

Abstract

Cloud data centers have become an important cyber-infrastructure vital to our society and economy, yet the virtualization infrastructure is prone to many reliability challenges. In a cloud, the virtualization infrastructure presents an abstraction layer on top of commodity hardware components and manages the execution of guest virtual machines (VMs). However, commodity computer systems are susceptible to hardware errors and expected to experience high error rates in the near future. Notably, a hardware error in a virtualized system can potentially result in a system crash, breaking the assumption of VM fault isolation. This reliability risk is further compounded by the massive scale of data centers hosting hundreds of thousands of servers as well as the pursuit of aggressive server consolidation that aims to host hundreds of VMs on each single server for high resource utilization. Traditional VM fault tolerant solutions are inadequate to address this challenge. This research proposes a hardware error resilient virtualization infrastructure that provides high-performance virtualization during error free execution while offering strong protection against hardware errors. This research will advance our understanding of hardware error behaviors in virtualized environments and develop virtualization-aware techniques that provide robust error detection, recovery, and protection. The proposed hardware error resilient virtualization infrastructure will offer a low-cost full system solution by taking advantage of the characteristics of virtualized systems and providing resource management mechanisms for balanced performance and reliability. The proposed virtualization infrastructure, if successful, will bring substantial benefits to cloud providers in delivering reliable services to millions of users. Education will also be an integral part of the project. This project will develop educational materials for the undergraduate and graduate curriculum, recruit and mentor under-represented students, and carry out a number of unique outreach activities that benefit K-12, undergraduate, and graduate students.