Efficient Cloud Storage Model for Service Optimization

Abstract

In a distributed data center enviornment, availability is the factor that cannot be compromised by the cloud vendors. Availability is achieved by data redun- dancy that involves replication across the data center. Replication is costly in data center as typically three replication results in thrice the resources. In order to provide redundancy, data service providers are increasingly mov- ing towards adopting erasure codes. There is a need to develop techniques that increase privacy of user data without increasing the complexity of the storage overhead in vogue. In the same context, we propose a novel cloud storage model that makes use of surplus storage and computing power avail- able on the user equipment to perform encoding and decoding of les along with providing an added location for storage of parts of les in addition to the datacenter. This is a step forward towards retaining user trust in the cloud and is inherently di erent from existing distributed computing solu- tions that use computational resources at any participating location. The proposed model uses users' own resources for coding and storage. Moreover, this makes it technically impossible for les to be stolen from the datacen- ter for any meaningful purpose without prior knowledge of the location and security of the user equipment, along with information about the coding scheme, which is highly unlikely. Results show performance comparison with conventional storage schemes that are in use today.