Vision: mClouds – Computing on Clouds of Mobile Devices

Vision

According to moor’s law -computation and storage will continue to increase. And this suggests that mobile devices in future, say in 2022 will have the same performance capacity as of our personal computers.  Though battery power and cellular bandwidth may not present similar trends but these future mobiles can still or will change the paradigm of mobile cloud computation. Authors present a vision of forming a cloud of such powerful mobile devices which are independent of backend clouds for computation and called it as mclouds. Mclouds vision is to explore the new hardware capabilities of mobile devices, making applications fast and reducing the cellular bandwidth usage –a growing network scalability problem. 

Key Ideas 

 

• First try doing it locally: Tasks which do not require too many resources can be accomplished by individual/personal processing such as PocketSphinx- a light weight speech recognition application for mobile devices. 
• Else form mloud:  Task can  be divided into sub-tasks and distributed processing technique can be deployed for performing computation. Device participating in mclouds are called mdevs and maintained as follows :
• Resource discovery: Find the mobile device willing to participate and store their unique identifiers.
• Formation:  Assign tasks  and start doing the computation.
• Maintenance: Make sure everyone successfully completes their job or else take help from backend cloud in cases  when devices invoke Release protocol i.e. ready to leave the mclouds. 
• Not working –contact backend cloud: If task requires too large amount of data or computation resources available in mclouds are not sufficient then computation can be moved to the main cloud. 

Challenges 

Mcloud approach relies on the user’s willingness to participate in formation of cloud. Thus proper incentives strategy should be inplaced for deploying mclouds in future. Paper explores the incentive of saving cellular bandwidth usage cost in different scenarios. Security and privacy still remains to be one of the main issue as compromising mobile devices are much easier then an enterprise cloud.

Final thoughts

These powerful mobile devices of future beg the following questions:

Will  such mobile devices make computation offloading schemes obsolete in future?

Do we need cloudlets anymore in future?