Vision: mClouds - Computing on Clouds of Mobile Devices

Motivation:

This paper provides us a vision of future that mobile devices as a core component of mobile cloud computing architectures. The authors present the vision of mClouds, which is a mobile cloud computing architecture that runs resource-intensive applications on collections of cooperating mobile devices, so that tasks can be accomplished locally without relying on costly and inefficient backend communication.

Main points:

The paper mainly talks about the mClouds design, especially mechanisms for managing the distributed computing operations between mDevs (mobile devices forming mCloud). Then the authors also give out some ideas about user incentive policies to drive mobile users use this technology. This paper illustrates mClouds in the following points:

• Design principle: local processing is better, so do it locally as you can. The reason is mobile devices will become powerful and task processing will need little or no support from cloud. So local processing can both syphon data off the cellular data channel and reduce the backend cloud complexity.

• Two ways: Processing tasks can either be executed on single device itself or distribute smaller portions to members in mCloud. By doing it individually, the tasks should not require too many resources. By distributing it to nearby devices, the tasks should be split into smaller subtasks with the conception of master mDevs and slave mDevs.

• Two important management principles: easily manage the mCloud and identify proper effective incentive mechanisms for users.

• Easily management has the following procedural steps: Resource discovery - master mDevs instantiate a distributed processing task and find nearby slave mDevs; Formation - slave mDevs join the tasks and tasks assigned; Maintenance - resource discovery phase executed several times to keep pace with possible topology change; Release - leave the mCloud.

• Incentive policies: would be helpful to encourage users lend their devices for others' computations.

Trade-offs:

• MClouds can take advantage of improving hardware capabilites on mobile devices.

• MClouds can reduce the impact on cellular data channel and then reduce bandwidth usage.

• Privacy and trust mechanisms and technologies still need to be discussed to encourage people's adoption of this technology and also make sure it is secure to use.