This project has been concluded from http://code.google.com/p/depsky/.
This project is set up to adapt depsky in such a manner, that it uses jclouds as a cloudprovider backend. During this process, we try to create a more user-friendly interface for easy embedded use in other programs.
We want to exclude some security measures, such as secret key sharing, since for our later use-case it is not necessary. Due to the remodelling of the system-architecture, we believe that it can easily be implemented later on.
- Driver usage unnecessary when using JClouds -> removed.
- New configuration protocol using yaml
- Code refactorings such as:
- Introducing new interfaces
- Getting rid of public access to variables
- Removing unnecessary code
- Main method not needed ( we do not intend to use this as a standalone client, rather than a library client )
- CloudManager fully relies on JClouds
- Reduced data in some objects
- CloudManagers are now callables and handle by an ExecutorService, instead of being Threads.
- Since JClouds gives the opportunity to differ between "blobs" and "containers" and also allows users to add metadata directly to a "blob", we
- Introduced a new paradigm, to retrieve metadata first and only if metadata is available, download (read) the payload. That leads to more distinctions in the operation set, such as DepSkyCloudMangager.GET_META.
- can now upload multiple dataunits to one container
The last changes regarding refactorings on the DataUnit side and also the handling of MetaData, leads to the renewal of the CloudManager workflow, aswell as re-programming the read and write operations on the client side. Since the metadata is added directly to the blobs, we have to change the DepSkyManager aswell.
This work released in the Apache License 2.0.
This project is supervised by Sebastian Graf from the Distributed System Group at the University of Konstanz.
- Sebastian Graf (Project Lead)
- Andreas Rain (Reprogramming & Redesign of the System)