RT-Push: A VM Fault Detector for Deadline-based Tasks in Cloud

Abstract

Nowadays, the cloud is becoming an important paradigm due to cost-efficiency, scalability, availability and high resource utilization. An increasing number of deadline-based (real-time) applications are moving to cloud for high availability and low latency. Applications like financial transactions, health-care system, scientific c work flows with their real-time nature, i.e., deadline bound, need provisioning of computing services, despite the presence of failure. So, cloud running deadline-based applications need a fault-tolerant framework that can assure availability and responsiveness. Fault can occur in any layer of cloud; physical, virtual, or application and can be handled by various fault management methods, which include fault avoidance, fault detection, and fault recovery. Here, our primary focus is on a Virtual Machine (VM) failure and its detection. We showed the VM behavior in normal working condition and with the failed state through state transition diagram. A timeout based VM fault detector RT-PUSH proposed for cloud running real-time applications. We used success ratio and execution drop rate as performance metrics to evaluate the effectiveness of the proposed fault detector.