Policy-based Management of Modular Cloud Storage Systems (Master Thesis, Finished)


Florian Lindörfer


Cloud storage systems are currently becoming more common. They promise large, scalable storage with simplified administration, low cost and only little risk, since the customer only pays for what she uses. Most cloud storage provider offer specialized solutions for specific scenarios and customers have to compare the properties of different solutions to decide on one that fits their requirements best.

For example, a customer may decide for a specific consistency protocol, but cannot express that his high level goal is a high consistency as far as it allows high throughput. To overcome this problem, we will propose a policy based management system for modular cloud storage systems. The policy system should allow the user to express its requirements independent of available modules in the modular cloud storage system. Given a policy, the cloud system should compose the necessary modules and try to optimize during runtime. Runtime optimization in this context deals with a changing environment in terms of usage statistics and also with a changing system as for example new modules. Furthermore, the system should be able to adapt to updates of the policies.

Since many scenarios require more than a single data-container type the customer should be able to define different policies for multiple container. Ideally the policies for on container could be defined independent of the other containers. Possible scenarios with multiple containers are a batch processing system with container for input and output data, or a multimedia archive with containers for multimedia data and meta data.

Download M.Sc. Thesis

Start / End Dates

2012/01/03 - 2012/07/02


Research Topics