EMFS: Email-based Personal Cloud Storage

Though a variety of cloud storage services have been offered recently, they have not yet provided users with transparent and cost-effective personal data storage. Services like Google Docs offer easy file access and sharing, but tie storage with internal data formats and specific applications. Meanwhile, services like Drop box offer general-purpose storage. Yet they have not been widely utilized, partly due to their fee-charging nature and long-term service availability concerns. Web-based email services, on the other hand, have been offering growing email storage capacity, reliable service, and powerful search capability, making them appealing as storage resources. In this paper, we examine the efficacy of leveraging web-based email services to build a personal storage cloud. We present EMFS, which aggregates back-end storage by establishing a RAID-like system on top of virtual email disks formed by email accounts. In particular, by replicating data across accounts from different service providers, highly available storage services can be constructed based on already reliable, cloud-based email storage. This paper discusses the design and implementation of EMFS, focusing on unique challenges and opportunities associated with utilizing email services for file transfer and storage, such as email based data organization, metadata format and management, and handling provider-imposed anti-spam usage restrictions. We evaluated EMFS extensively with multiple benchmarks, and compared its performance with NFS, AFS, and a non-free cloud storage service built upon Amazon S3. Our results indicate that while EMFS cannot match the performance of highly optimized distributed file systems with dedicated servers, it performs quite closely to the commercial cloud storage solution.