Observed Availability of Cloud Services

Cloud computing is used widely and is going to be used even more in the future. Many internet-based services are now designed to be 'cloud native' using architectures that allow them to take advantage of the scalability of the underlying cloud infrastructure allowing customer services to meet potentially rapid changes in customer demand. While there are many customers successfully leveraging cloud services for their bene t, the use cloud computing has also drawn critique on its other aspects such as its reliability and security. Although issues of security and operational cost benefits have been studied and actively marketed by the major cloud infrastructure service vendors, the question of service reliability and more specifically, availability of cloud services is less researched in academia and also less publicised by the cloud vendors themselves. This study takes a look at the service availability of cloud infrastructure services. The study focuses on the largest public cloud infrastructure provider e.g. Amazon Web Services, and uses publicly available incident information to analyse outages from multiple viewpoints. The use of publicly available information at has allowed this work to analyse a wider selection of services than earlier studies, but also does limit the scope of outages that can be analysed to relatively large-scale outages. The overall result is that Amazon Web Services' services during the analysis period of June 5th 2013 to June 4th 2014 are reliable services with an overall availability of 99.983% over all of the services included in this study. During the analysis period there was a total of 139 separate outage events where an average outage event lasted 130 ± 20 minutes. EC2 and RDS, two of the services with known Service Level Agreement availability target, meet their contractual availability targets by a comfortable margin with both having over 99.9999% availability when measured in comparable units to the SLA's target of 99.95% availability