Cloud Oriented File Service for Mobile Internet Devices

Along with the rapid growth of heterogeneous cloud services and network technologies, more mobile devices use cloud storage services to enlarge the capacity and share data in our daily lives. We commonly use cloud storage service clients in a straight forward fashion, since we may easily obtain most client-side software from many services providers. However, when more devices and users participate in heterogeneous services, the difficulty increases to manage these services efficiently and conveniently. In this paper we design and implement a novel cloud-oriented file service, Wukong, which provides a user-friendly and highly-available facilitative data access method for mobile devices in cloud settings. By using the innovative storage abstraction layer and a set of optimization strategies, Wukong supports heterogeneous services with a relatively high performance. By evaluating a prototype in a systematic way on the aspects of the supporting interface, system performance, and the system resource cost, we find that this easily operable file service has a high usability and extensibility. It costs about 50 to 150 lines of code to implement a new backend service supporting plugin. Wukong achieves an acceptable throughput of 179.11 KB/s in an ADSL environment and 80.68 KB/s under a country EVDO 3G network with negligible overhead.

The rising and demanding use of mobile devices like smart phones, Netbooks, iPads and Mobile Internet Devices (MID) [1] grows steadily in daily life. Many people commonly operate several devices at the same time, including laptops, phones, digital picture frames, music players, and so on, often constrained by limited storage capacity. One commonly used method overcomes this problem by copying the data not used soon by a backup device or service (i.e., Amazon S3 [2], Google Docs [3], Dropbox [11]). When we subject changes to the data, then the user synchronizes their data on each device, one by one. Thus, manual migration and synchronization of data between different places happens frequently. Meanwhile, mobile device users also commonly exchange information to collaborate and share data. The demands of our daily actions, like data copying and sharing across multiple users, use resources ubiquitously. With the development of the cloud computing [4] and network technologies, we have met potential feasible methods for large storage capacity and convenience of sharing. Mobile devices carry services, like Google Docs, Amazon S3, and MobileMe [5] as the storage backend in order to enlarge the capacity of storage, and achieve resource sharing via mobile network ubiquitously. However, difficulty increases when more devices and users participate in and more applications need to use the heterogeneous resources in cloud storage to keep their world in sync and convenient. Four challenging issues arise in using such systems for mobile devices.
1) Most of the mobile applications prefer or only support
local file system interfaces, so a mobile user has to
download files from the service before using them
locally. But for mobile devices, especially for those
with poor user interface [6, 19], finishing an action with
complex operations creates a terrible experience.
2) The users always use multiple types of services in order
to satisfy their common need. A system that supports
only a special service would not suit our users to access
resources conveniently. Since different service
providers always offer different public APIs, users
should pay attention to the service that they are using
but not transparently use the services.
3) With many types of network, the latency and the
network bandwidth vary with network types. In order to
overcome these challenges, the system should consider
additional optimization in the design.

Free download research paper