Availability in the cloud: State of the art

Availability is a non-functional requirement defined as the percentage of time a system or a service is accessible. This percentage determines the acceptable total outage time for any given period. High-Availability (HA) is a stringent requirement which allows for a maximum of approximately five minutes downtime in a year including outage due to scheduled maintenance and upgrades. It is agreed that availability is among the main challenges of the cloud. There has been a lot of work on availability in cloud computing, but cloud providers and researchers from the academia and the industry have used different definitions for availability and the related concepts. Thus, it is difficult to evaluate and compare different solutions. In this paper, we present a survey of availability solutions proposed for the cloud by the main cloud providers and by 21 most relevant conference and journal papers out of the 100 papers collected initially. To conduct this survey we defined a taxonomy, which captured the main concepts, mechanisms and metrics for availability. We use this taxonomy to evaluate and classify the solutions of cloud providers as well as solutions proposed in research papers, their strengths and weaknesses. We point out potential future research directions.