|
Abstract:
|
"In order to address reliability and availability of cloud services, redundancy-based techniques utilizing more than one cloud provider have been recently proposed. Unfortunately these approaches fail to recognize the effects related to common dependencies hidden by in- dependent redundant services, potentially invalidating these efforts. We propose a novel system to address this pitfall by recommending cloud consumers the most suit- able redundant services based on the consumers’ requirements. We call this system a cloud reliability recommender (CRR). At the heart of a CRR, we leverage fault tree analysis techniques to 1) discover hidden common dependencies in order to generate correlation matrices; 2) calculate the failure probabilities of alternative services; and, 3) enable cloud consumers to specify criteria to optimize their utility, recommending the most suitable services for them. In this paper, we describe the CRR, the aforementioned process, and discuss the challenges inherent to CRR design and practicality."
|