| Along with the rapid development of science and technology,on-going mobile devices have become more and more important in people's life.These types of devices quickly become popular,based on the ability of mobile devices to access information anytime,anywhere.Nowadays,mobile devices have become a research hotspot and achieved great success.However,the development of mobile devices has also encountered some difficulties.Limited constrained mobile device still face a big challenge in related fields such as communication and computation.In recent years,the ratio of mobile devices is growing progressively.More and more web applications are change into the mobile application because it is small in size and convenient to carry.But the mobile devices are facing challenges like limited storage and processing capability in key areas such as communication and computational.Since real-time applications like Augmented Reality,Video Analytics,and 3D-Games are comprised of computing and dataintensive tasks.It may require powerful processing machines and large-scale data storage against limited constraint to guarantee the real-time application.To get rid of this situation,the mobile cloud computing(MCC)architecture has partitioned the application execution into local parts,and remote parts(i.e.,cloud computing).The ultimate goal of the MCC is to augment the competency of the mobile device and minimize the system energy consumption.The fundamental job of the offloading method is to divide application execution into local and remote execution.The computational offloading to the remote cloud via a wireless system introduces communication cost.However,the foremost cost of the offloading system is the combination of computation and communication time for the whole process.While mobile cloud computing makes a great contribution to our daily lives,it will also,however,bring numerous challenges and problems.In short,the core of such challenges and problems is just how to combine the two technologies seamlessly.On one hand,to ensure that mobile devices make the best use of the advantages of cloud computing to improve and extend their functions.In this study,we focus on task offloading,which must be energy efficient and robust.We are studying the dynamic application partitioning and an energy efficient task assignment in MCC.The considered problem has two costs such as communication and computation while performing offloading.The main objective of this thesis is to minimize the total energy consumption of the system,i.e.,local device energy and communication energy in.The offloading is designed to save energy for the mobile devices.We have divided the considered problem into application partitioning and task scheduling.The flowchart of the application partitioning starts with profiling and static analysis technology.The application light tasks which are executed locally and compute-intensive tasks are offloaded to the cloud server for further performing by resource managers.If the resources are enough for tasks to execute on the mobile device,it may be performed locally.Otherwise,the offloading system will offload tasks to the cloud-based factors,for example,the task size,the available network bandwidth,the cloud speed and the mobile processing speed.Make sure that this partition could not be higher than the mobile power threshold.Application solving and application partitioning would be done dynamically during the time,and synchronous task migration is being processed.The offloading system process runs till all tasks are mapped by an appropriate cloud server in order to minimize total energy consumption.In our work,we represent the Application as a directed acyclic graph,i.e.,G(V,E),the tasks can be shown by V.E is a communication cost between two adjacent tasks.The application tasks always follow the topological order that is a call graph.After partitioning,the call graph is converted into a consumption-weighted graph.In this thesis,we investigate the energy consumption in the mobile cloud architecture.Applications such as voice recognition,natural language processing and video conferencing require higher CPU instruction clock frequency and consume more energy.Generally,the energy consumption in the mobile cloud architecture can be determined by the local device energy model,communication energy model,and cloud execution energy model.However,in this thesis,we just focus on mobile energy consumption execution model.We only optimize mobile power consumption model.Transmission energy consumption model and the cloud execution model are not considered in our study.To minimize the energy consumption of the mobile device,the technique offloads the relative compute-intensive tasks from the closely related proximity cloud server;thereby it prolongs mobile battery by the offloading system,which may save the energy consumed by CPUs on the mobile device while accomplishing divergent applications.We recommend the Asymptotic Time Complexity method to differentiate the computational complexity of the divergent applications on mobile devices while executing their process.We have proposed EETA(energy efficient task assignment)algorithm that is a partial offloading technique so that minimizes and prolongs the device energy as well as communication energy.Application processing time has a linear relationship with power consumption.Our proposed EETA executes all tasks within polynomial time.Generally,the proposed algorithm is iterative in nature.However,the more fundamental complexity is the time complexity function for EETA algorithm,which is We know that EETA is iterative in nature,so in this case,V is the number of iteration for either assigning the tasks locally or remotely in order in order to minimize power consumption.And E is the time complexity of finding the optimal cloud and mobile resources which incurs the minimum response time in each iteration DVFS is a technique able to monitor the application power consumption during execution life the cycle.The application could be scaled up and scaled down due to many factors such as heterogeneity in communication technologies,cloud resources,and mobile devices.So it is not trivial to execute the same application with a similar rate of power consumption.There are many features compacted to the mobile device application such as mobile device mobility,it will consumes much more energy as communication technologies change from the cellular network to the Wi-Fi or LTE connection.In the end,the proposed EETA is reliable and efficient in the adaptive environment in which networking connection,cloud resources,and mobile status adaptively change during the application execution life cycle.For the considered problem,we have proposed an Energy-Efficient task assignment algorithm.The preliminary objective is to minimize the mobile energy consumption,whereas mobile energy consumption has a linear relationship with communication energy and process time at the cloud server.Generally,the application has dependent tasks that follow precedence order.Naturally,the application can be illustrated as a call graph.After application partitioning,the call graph is converted into a consumption graph.And the consumption graph has two sets of tasks and represented as disjoint sets(i.e.,local execution tasks and remote execution tasks).The complexity and fascination of the proposed algorithm must meet the polynomial time.Offloaded tasks are transported by virtue of communication protocol.In conclusion,the proposed algorithm EETA outperforms comparing to baseline offloading approaches.The considered application tasks can be shown via call graph.We can predict the application execution time via profiling technology.However,mobile device offloads the compute-intensive tasks to the cloud server via wireless technology.We can predict wireless bandwidth via network profiling technology.Whereas,existing offloading studies focused on stable mobile cloud environment(i.e.,fixed bandwidth and cloud resources),but it is unrealistic where network and cloud server values change over time.Therefore,we take offloading during runtime.In the meantime,compared to existing methods,it can adapt to changes in the runtime environment.To evaluate the EETA efficiency in the context of energy efficiency three kinds of calibration values such as unchanging standards: some set of parameters is fixed by the application developer during design,analysis such as power conservative elements which are required to be fixed for different kinds of devices.We have examined several real-time application workloads such as a Linpack math tool,3D-Game EEG Beam 3D Game,Augmented Reality(a gesture application)and Face Recognition application.Each workload is characterized by some properties such as task size,memory requirements,required power consumption,and execution time.The individual application has benchmark application features,and each task is required to be executed within a given threshold which could be energy or execution time bound.We show the application via Call graphs and it is basic program analysis results that can be used for human understanding of programs,or as a basis for further analyses,such as an analysis that tracks the flow of values between procedures.One simple application of call graphs is finding procedures that are never called.Call graphs can be dynamic or static.A dynamic call graph is a record of an execution of the program,for example as output by a profiler.Thus,a dynamic call graph can be exact,but only the program running once.A static call graph is a call graph intended to represent every possible run of the program.The exact static call graph is undeniable,so static call graph algorithms are generally over-approximations.That is every call relationship that occurs is presented in the graph,and possibly also some call relationships that would never occur in actual runs of the programWe compared the proposed energy efficient algorithm EETA with baseline approaches,For instance,non-offloading.Full-offloading Extensive simulations have evaluated the performance and efficiency of the algorithm based on different instants,for suppose power consumption on interactive applications,benchmark application,augmented reality application and face recognition application.The proposed approach is the linear approach,whereas process time has a direct relationship with energy consumption.In this thesis,we are a formulating a task offloading problem.The objective is to execute real-time applications likewise Augmented Reality,Healthcare,3D-Game and etc.on a resource constraint mobile device.However,the mobile cloud computing architecture is a promising method to allow the mobile device to execute compute application in a collaborative manner with cloud computing via offloading technique.There are two kinds of offloading technique such as code offloading and task offloading.In general,code offloading performs data offloading to the cloud.Server task offloading performs compute intensive from a mobile device to the cloud server.In this thesis,we are focusing on a task offloading technique.The study proposes a novel Energy-Efficient Task Offloading(EETA)algorithm to determine the optimal assignment of application workload to the appropriate cloud computing in order to minimize the energy consumption of a mobile device.EETA calculates two-level energy costs such as local cost and communication cost,local cost determines the execution of tasks on the mobile resources and consumes the battery power,and compute-intensive tasks offloaded to the remote cloud via wireless network it incurs communication cost.To find the effectiveness and efficiency of the EETA we compare with baseline offloading techniques likewise non-offloading and full offloading.There are four workloads deployed for experiment setups to evaluate the effectiveness of the proposed EETA algorithm.Mobile Cloud Computing is gaining its momentum and is encouraging the users to leverage the benefits that Cloud provides in order to ease the burden on resource constraint mobile devices.Offloading can become beneficial for computationally intensive applications,provided entities impacting the offloading are understood and analyzed thoroughly.Future research work is needed in developing efficient Offloading frameworks and optimal partitioning schemes which are portable across all the mobile execution platforms.The EETA framework proposed in this thesis about the power consumption on Idle and waiting time for task result and disk usage.Dynamic adaption could happen during the application execution lifecycle I.e.,network and device profiling change by over time.The preceding condition produces in the accurate offloading result.In this thesis,we study the power management during the offloading while minimizes total cost as much as possible.Proposed power efficient task workload assignment(EETA)Energy Efficient Task Assignment algorithm dynamically partitions the execution of application into two sides(i.e.,local execution and remote execution),which overcome the shortcomings of previous algorithms.EETA has certain steps.First,analyze and profile the application,which executed either locally or offloaded to the remote cloud,second,if the weight of the application is more than processing power of the mobile device,then partition the runtime and maps them with a resource on both sides(e.g.,mobile execution,and cloud execution).Based on the greedy algorithm,the EETA is iterative in nature;it always chooses optimal resources for offloaded tasks in order to minimize the total execution time.The evaluation shows that EETA augments the system performance with the adaptive environment and dynamically optimize the performance as compared to baseline approaches.The final outcomes can be seen that EETA reduces energy consumption by 50%.This paper focuses on how to manage energy during offloading and how to reduce total cost while managing energy.Main tasks as follows:1.To partition the application dynamically and effectively in an adaptive environment,we have proposed the EETA(Energy Efficient Task Assignment)algorithm.EETA has three phases,application partitioning phase: this process can be done based on available parameters,i.e.,mobile CPU speed,storage,available bandwidth,and server speed.Previous baseline approaches used static partitioning that could be failed while variation occurs in the above parameters.Second phase: network contexts(i.e.,bandwidth)can be changed during different interval.The EETA was proposed to get rid of this situation for re-partition of the application according to newly available parameters.Third phase: task scheduling at the local and remote cloud in order to minimize the total system energy consumption(e.g.,mobile and communication energy).2.The proposed algorithm EETA is greedy based on which iteratively choose the optimal offloading system either task offload or not so that minimize the cost.3.The simulation results show that our proposed EETA reduces the energy of the whole system significantly as compared to baseline approaches such as non-offloading and full offloading.We find out the efficiency and effectiveness of EETA on benchmark applications(e.g.,face recognition application and EEG Beam Game).The EETA notably reduces the energy consumption while performing offloading,statistically,the results shows the implication of EETA as compared to baseline strategies. |
PDF Full Text Request