GIT-CERCS-15-02
Kyle Bilbray, Douglas Sigelbaum, Douglas M. Blough,
GeoShare: Experience with a Geographically Diverse Cloud Data Storage Service
In this report, we describe our experiences in developing GeoShare, a
geographically diverse cloud data storage service. Users are becoming
increasingly wary of storing their sensitive data within servers physically
located in one country or within one political or legal boundary. GeoShare
facilitates the fragmentation of information across administrative and
geopolitical boundaries, thereby maintaining data confidentiality from powerful
adversaries such as cloud administrators and nations. Our approach includes a
method to calculate placements of information fragments to satisfy users'
geographic constraints and simultaneously optimize system performance. We also
present an approach to optimize cost/performance by considering the storage
cost versus latency tradeoff that arises from replicating data across far-flung
geographic regions. Our approaches have been fully implemented, tested, and
evaluated in the GeoShare prototype cloud storage service, which runs on top of
Amazon Web services and makes use of 8 world-wide regions accessible through
the Amazon S3 cloud storage service. We demonstrate the system through
experiments having Amazon EC2 clients all over the world simultaneously
creating, uploading, and downloading objects using the GeoShare service.