The Zebra striped network file system
ACM Transactions on Computer Systems (TOCS)
SOSP '03 Proceedings of the nineteenth ACM symposium on Operating systems principles
Ceph: a scalable, high-performance distributed file system
OSDI '06 Proceedings of the 7th USENIX Symposium on Operating Systems Design and Implementation - Volume 7
Computer Networks: The International Journal of Computer and Telecommunications Networking
The case for enterprise-ready virtual private clouds
HotCloud'09 Proceedings of the 2009 conference on Hot topics in cloud computing
Volley: automated data placement for geo-distributed cloud services
NSDI'10 Proceedings of the 7th USENIX conference on Networked systems design and implementation
Secure cloud computing with a virtualized network infrastructure
HotCloud'10 Proceedings of the 2nd USENIX conference on Hot topics in cloud computing
Hi-index | 0.00 |
An enterprise uses VPNs, leased from a service provider, to interconnect multiple sites that are geographically apart. The service providers, as they start providing cloud-based services, are finding themselves well-positioned to providing storage services in the cloud for an enterprise, and make the service accessible through the existing VPN connections. Enterprise users, however, are used to fast, ubiquitous and guaranteed access to the storage from any enterprise location. This is achieved by having network attached storage (NAS) connected to the enterprise network. In order to maintain the same level of service, when the enterprise storage is moved into the cloud, the service provider must ensure that the storage is accessible from all the enterprise locations as if it is connected to the enterprise network itself, regardless of the actual user or the file. In this paper, we present a system that enables cloud storage service with guaranteed performance from all published access locations of an enterprise. Knowing only the limits on users access rates or their access bandwidth limitations, we develop an access oblivious storage provisioning and placement strategy. Our system uses a combination of chunking, data replication and intelligent data placement to guarantee performance to accessing the storage in an access independent manner without significant over-provisioning.