Mobile devices have become the major computing platform in todays world. However, some apps on mobile devices still suffer from insufficient computing and energy resources. A key solution is to offload resource-demanding computing tasks from mobile devices to the cloud. This leads to a scenario where computing tasks in the same application run concurrently on both the mobile device and the cloud.
This dissertation aims to ensure that the tasks in a mobile app that employs offloading can access and share files concurrently on the mobile and the cloud in a manner that is efficient, consistent, and transparent to locations. Existing distributed file systems and network file systems do not satisfy these requirements. Furthermore, current offloading platforms either do not support efficient file access for offloaded tasks or do not offload tasks with file accesses.
The first part of the dissertation addresses this issue by designing and implementing an application-level file system named Overlay File System (OFS). OFS assumes a cloud surrogate is paired with each mobile device for task and storage offloading. To achieve high efficiency, OFS maintains and buffers local copies of data sets on both the surrogate and the mobile device. OFS ensures consistency and guarantees that all the reads get the latest data. To effectively reduce the network traffic and the execution delay, OFS uses a delayed-update mechanism, which combines write-invalidate and write-update policies. To guarantee location transparency, OFS creates a unified view of file data.
The research tests OFS on Android OS with a real mobile application and real mobile user traces. Extensive experiments show that OFS can effectively support consistent file accesses from computation tasks, no matter where they run. In addition, OFS can effectively reduce both file access latency and network traffic incurred by file accesses.
While OFS allows offloaded tasks to access the required files in a consistent and transparent manner, file accesses by offloaded tasks can be further improved. Instead of retrieving the required files from its associated mobile device, a surrogate can discover and retrieve identical or similar file(s) from the surrogates belonging to other users to meet its needs. This is based on two observations: 1) multiple users have the same or similar files, e.g., shared files or images/videos of same object; 2) the need for a certain file content in mobile apps can usually be described by context features of the content, e.g., location, objects in an image, etc.; thus, any file with the required context features can be used to satisfy the need. Since files may be retrieved from surrogates, this solution improves latency and saves wireless bandwidth and power on mobile devices.
The second part of the dissertation proposes and develops a Context-Aware File Discovery Service (CAFDS) that implements the idea described above. CAFDS uses a self-organizing map and k-means clustering to classify files into file groups based on file contexts. It then uses an enhanced decision tree to locate and retrieve files based on the file contexts defined by apps. To support diverse file discovery demands from various mobile apps, CAFDS allows apps to add new file contexts and to update existing file contexts dynamically, without affecting the discovery process.
To evaluate the effectiveness of CAFDS, the research has implemented a prototype on Android and Linux. The performance of CAFDS was tested against Chord, a DHT based lookup scheme, and SPOON, a P2P file sharing system. The experiments show that CAFDS provides lower end-to-end latency for file search than Chord and SPOON, while providing similar scalability to Chord.
|
![[NJIT logo]](http://archives.njit.edu/vhlib/images/njit_logo.gif){kind=logo}
![[Van Houten Library logo]](images/vhl-top-image.jpg)
