Data Process Tasks Placement In A Cloud-Edge System

With the development of the Internet of Things(IoT),more and more IoT appli-cations are deployed.IoT devices often generate a large amount of real-time stream-ing data that needs to be processed by IoT applications.In practice,the data streams that an application needs to process come from a multitude of geographically diverse devices.These data streams are numerous,consume a lot of bandwidth,and geo-graphically diverse.The WAN bandwidth is limited and end-to-end latency is high,which is likely to cause delays and even data loss.In order to solve the contradiction between the WAN limitations and application's low latency and high transmission rate requirements,researchers in related fields have proposed computing systems and modes called Edge Computing(or Fog Computing).By deploying a small computing server close to the user's network edge,such as a wireless base station,an edge local area network,etc.,data processing tasks are deployed therein to satisfy the applica-tion's data processing requests,saving time for data transmission to the remote data center.Deploying streaming data processing tasks under the edge computing framework,the ensuing question is which strategy to use to deploy data processing tasks.Due to the difference in delay of multi-input data streams with different geographical distri-butions and the competition of bandwidth resources between multiple applications,different application deployment locations may have different application delays and bandwidth resource competition effects.At the same time,when the system can allow multiple tasks exiting in one task,the task location and data flow routing will affect each other during the placement of multiple copies of a task,making the problem more complicated.In order to solve the above problems,the streaming data pro-cessing task can be efficiently operated in the multi-edge node-data center computing system.This paper studies the corresponding non-copy and multi-copy streaming data processing task placement problems.Firstly,this paper studies the problem of task deployment without replicas and gives an algorithm strategy to efficiently solve the task deployment problem without replicas.Under the constraint of the node bandwidth in the edge network environment,the relationship between the bandwidth consumption of the system computing node and the task deployment location during the task deployment process is described by the streaming data task model,and the data stream end-to-end is minimized.Deferred as a goal,formal description and analysis of task-free deployment problems without replicas.This paper designs a global node collaborative sensing algorithm strategy to solve the problem of task deployment without replicas.In the scenario where more generally data processing tasks can be paralleled,this paper also studies the task deployment with replicas problem.In this paper,the task deployment problem of multiple copies is decomposed into two aspects:task deploy-ment and data stream routing.Then,through the strategy of progressive solution,the problem of multi-copy task deployment is solved.In this paper,the efficiency of the proposed method is verified by simulation ex-periments.Compared with the algorithm that can't perceive the multi-edge node bandwidth resource competition and the delay difference between edge nodes,the proposed algorithm can achieve lower data stream processing delay and higher band-width usage.