Lowering the Barriers to Large-Scale Mobile Crowdsensing

Motivations:

Deploying mobile crowdsensing applications in large-scale environments creates a a tremendous burden on application developers as well as mobile users. There are three major barriers to achieving the large crowd sizes critical to their success: heterogeneity of sensing hardware and mobile platforms, too much applications burden on users and increasing network bandwidth demands of applications.

Main Points: 

The author introduce his method by 3 parts: principles, architecture and deployment. 

Principles:

In order to solve the three problems, the author proposes three principle to create deployment model.

• Separate data collection and sharing from application specific logic. By using this method, sharing data process runs in Virtual Machine, which solves the problem.
• Remove the installation on smartphone, which reduce the burden of users. 
• Decentralize of processing and deal with data in the server near the source of data. 

Architecture:

Based on the 3 principles above, the paper propose a  3-tier system architecture. The first level contains mobile devices. The second level is a distributed cloud infrastructure. The mobile devices level collect data and send the data to the nearest server, which runs on the second level. A master  application, which runs on the third level, controls the second level application.


Deployment:


Deployment is a process to request permission to generate application by cloudlet daemon. This process based on two assumptions: the architecture relies on distributed cloud near the user and there exists a way to transfer data between two VMs.

Trade-off

• The novel idea of this paper is that, by separating data collection and sharing,  users do not need to install too many apps on their smartphone and developer do not need to face different platforms. 
•  The paper claims that the model can reduce the traffic on wide-area networks or latency. But the author does not give any evaluation about this benefit. 
• This paper does not introduce any potential security problem of the model.