Contextualization: Dynamic Configuration of Virtual Machines
(English)Manuscript (preprint) (Other academic)
Virtual Machines (VMs) are commonly used as building blocks of IaaS cloud services. The number of running VM instances can be adjusted during runtime to achieve elasticity in the capacity of the service. New VM instances are based on templates that contain the basic configuration of the VM. Instance specific settings, settings unique to the infrastructure to which the instance is being deployed, are normally injected to the VM during the deployment phase through means of contextualization. In this work we present a layered approach to contextualization that supports different classes of contextualization data through the use of virtual devices. Inside each VM, data from different classes are layered on top of each other to create a unified file hierarchy using a small, custom file system. Context data can be updated during runtime by updating the contents of the virtual devices, making this approach the first contextualization approach to natively support recontextualization. Recontextualization enables run-time reconfiguration of a running service and can act as a trigger and key enabler of self-* techniques running inside the VM, allowing the service itself an unambiguous trigger for, e.g., further optimization in response to a changing environment. The runtime reconfiguration using recontextualization and its potential gains are shown in an example with a distributed file system, demonstrating the feasibility of the approach.
Cloud Computing; Contextualization, Recontextualization, Configuration, Virtual Machine
Research subject Computer Science
IdentifiersURN: urn:nbn:se:umu:diva-80128OAI: oai:DiVA.org:umu-80128DiVA: diva2:646918
FunderEU, FP7, Seventh Framework Programme, 257115
Submitted 20132013-09-102013-09-102013-09-19Bibliographically approved