Research Related To Task Allocation Framework And Algorithm In Mobile Cloudlet

With the rapid development of cloud computing,wireless communication,smart mobile devices and mobile cloud computing,new technologies are emerging constantly.Mobile cloud computing is applied in more and more aspects of our real life.People do not have to wave to taxi drivers to call a taxi but use applications in smart phone to send a request to all drivers within few miles.When we are tried on the street,we don't have to walk to find a restaurant but use applications to find a restaurant which do not require a long walk to be there and with high reputation.People do not have to daze in the bus or underground but read some news recommended by the systems based on our interest or watch the latest TV series.Although the computation capability of our smart phones increase a lot,these devices still cannot finish computation-intensive tasks very fast like servers in the data center or computers at home.Even if sometimes users do not cares about response time,computation-intensive tasks will consume great amount of battery power which lead to low battery stand-by time.These two drawbacks of smart phones hinder some time intensive applications like augmented reality(AR)or games running locally.In order to run these computation-intensive applications on smart phones,one solution is to offload these applications to the cloudlet.A cloudlet is a mobility-enhanced small-scale cloud datacenter that is located at the edge of the Internet.The main purpose of the cloudlet is supporting resource-intensive and interactive mobile applications by providing powerful computing resources to mobile devices with lower latency.Compared with traditional method that offload to remote data center,there is no such problems like latency caused by congestions or jitters in the network or long distance in geography.The drawback of cloudlet is that in most cases,there isn't great number of nodes in the cloudlet,furthermore some of these nodes are mobile nodes and use battery as its power source.These drawbacks will lead to two problems.One is that nodes may leave the cloudlet while working on some tasks from mobile clients or helping to transmit some task data.The other is that the power draining rate of batteries maybe very unbalance.Nodes may leave the cloudlet one by one and finally there maybe insufficient nodes to finish the tasks or some mobile clients may fails to connect to the cloudlet.The aim of this paper is to reduce the impact of these two problem by designing a new service process and allocating algorithm.The main idea is to avoid using cloudlet nodes with high mobility and trying to balance the power draining rate while allocating tasks.To be more specific,mobile clients needs to upload request first and transmit task code and data after cloudlet replies where to upload.The allocating algorithm will take battery power,mobility,computation capability of each node,network topology into account and find a suitable node and the route from mobile client to this node.Simulation result shows that the time that cloudlet with our service process and allocating algorithm could provide QoS is much longer.In the last section,we design a new centrality definition which is more suitable compared with existing centralities to evaluate the importance of each nodes in the cloudlet based on network topology.Simulation results shows that this evaluation method will lead to longer time to provide QoS for cloudlet.