Research On Content Caching And Task Migration Algorithms Based On Cloud-Edge Collaboration

With the diversification and complexity of business types in the Internet of Things,the traditional cloud computing model is facing severe challenges such as weak real-time performance and high bandwidth pressure.Mobile edge computing(MEC)with sinking computing power can reduce the bandwidth pressure while providing low latency services,which is especially suitable for delay-sensitive,bandwidth-intensive and other emerging businesses.However,the limited computing resources make MEC servers prone to overload in high-concurrency or computing-intensive scenarios,which will lead to task blocking or even server crashes.The abundant resources of cloud computing can just provide assistance for MEC to complement each other.Therefore,this thesis focuses on the overload of MEC servers under the cloud-edge collaboration architecture.Specifically,this thesis starts from the two perspectives of "overload prevention" and "handling overload",and focuses on the two key technologies of "content caching" and "task migration".The main work involved in the research is as follows:Firstly,the content caching technology mainly consumes storage resources in exchange for reducing the computing load.However,the existing research has not fully considered the characteristics closely related to the computing load of the MEC server,such as many computing-intensive tasks and strong content timeliness in the MEC scenario.Although it can reduce a small amount of computing load on the MEC server,it cannot effectively prevent overload.Therefore,this thesis analyzes various factors related to the "computation" and "storage" of the MEC server in combination with the characteristics of the MEC scenario,and then designs a cache value calculation method that considers multiple factors.Finally,a cloud-edge collaborative content caching algorithm based on the cache value is proposed.The simulation results show that compared with the MCCR algorithm,the effective cache hit rate of this algorithm can be improved by up to 20.08% and the task blocking rate can be reduced by up to 12.53%,which can effectively prevent MEC server overload while guaranteeing the system cache performance.Secondly,the task migration technology reduces the load pressure on the MEC server by transferring the load to cloud center.This thesis focuses on the fact that the actual load of MEC servers is usually dynamically fluctuating and unbalanced.However,most of the existing research is based on the global load of the MEC layer for task migration,which cannot fully consider the load differences between MEC servers and cannot accurately capture the changes in the load of MEC servers.So,it cannot deal with overload in a timely and effective manner.In this regard,this thesis designs a task migration algorithm at the server granularity that can carry out overload protection adaptively.The algorithm can be divided into three parts: load monitoring,task admission,and task migration.The task admission mechanism is the basis for task migration decisions,which adjusts in real-time based on the results of load monitoring.The real-time load monitoring is related to the load fluctuation of the MEC server,and then adaptive overload protection can be realized according to the load fluctuation.Finally,this thesis compares the proposed algorithm with the traditional task migration algorithm based on global load.The experimental results show that the task blocking rate of this algorithm can be reduced by up to 36.9%,which can adapt to load fluctuation for real-time overload protection.