Dependability Differentiation in Cloud Services

Cloud computing is an evolving computing paradigm in which computing is provided to consumers upon request as a service and is charged on pay-as-you-go basis. The virtually infinite computing resources on the cloud provider side and the dramatic decrease of investment requirement on the consumer side have made large companies to consider outsourcing their IT services to cloud providers. But, large companies have very strict dependability requirements if they have to outsource their internal IT functions to clouds. Their requirements differ largely from a domestic cloud user. The cloud providers are, thus, supposed to provide different levels of dependability to different types of users depending upon the SLAs between them. One of the dependability attributes for cloud services is service availability which is the availability perceived by a cloud user. It is an umbrella attribute which gathers cloud availability, network availability, cloud performance, network performance and cloud security under it. It is also one of the most important dependability requirements of most of the large IT functions. In the project work, autumn 2010, analytical models for differentiating cloud availability by replicating Virtual Machines (VMs) were developed, using different replication schemes. This thesis work will build upon these models, refining them where necessary, and analyzing the achieved dependability differentiation. In addition to that, investigations to find other techniques for dependability differentiation both on the cloud and the network side (both availability and performance) will be made. Service availability will then be accounted for as a whole and comparisons will be made to find out that on which part of the cloud service provision (cloud or network) it is worth investing in terms of resources and finance to achieve better differentiation. The following tasks will be carried out in thesis: 1. Investigate and identify possible techniques for dependability differentiation both in clouds and networks. 2. Identify example services with different dependability requirements: a. Single service with different dependability requirements for different user classes b. Different services with different dependability requirements 3. Examine the models from the project work with respect to simplifications and realism 4. Develop models for a selected set of scenarios with differentiated dependability in both clouds and networks for example service(s). 5. Simulate the scenarios and compare them. 6. Discuss which part of cloud service provision (cloud or network) is worth investing in, in terms of resources and finance, to achieve better dependability differentiation. Abstract As cloud computing is …