NCU-HA: A Lightweight HA System for Kernel-Based Virtual Machine
Yen-Lin Lee, Min-Huang Ho, Aswin Suharsono, Yu-Chen Pan, Wei-Jen Wang, Deron Liang · 2017
Failure prevention and protection are very critical issues nowadays. Although there are many commercial high availability (HA) solutions, they are expensive and do not support intelligent platform management interface (IPMI) that can improves failure detection efficiency. We present NCU-HA, a HA solution based on the open source project KVM(Kernel-based Virtual Machine). NCU-HA can be run on both low-cost personal computers (PCs) and IPMI-based servers. Moreover, to prevent single point of failure, we classify the nodes of the cluster into three roles: Primary node, Backup node, and General nodes. HA agent on those nodes will provide the functions depending on its role. We also define failure models and describe the recovery methods implemented in NCU-HA. Evaluation results show that our system can reach four nines (>99.99 %) availability on low cost PCs.