论文阅读报告

Q1: What problem does this paper try to solve?

This is the statement that I personally think is more appropriate. The statement in this paper is \"It may be necessary to offload the computation to a higher-resourced compute node in the cloud.\

Deploy lambda functions in cloud and edge with different configuration.

Conducts the minimization algorithm to select the configuration for the input.

   This framework targets intelligent applications consisting of a single serverless function that excutes a data processing task on an input.

I checked the brief description on the AWS website (I didn't have enough time to read the detailed docs) and saved the relevant sreenshots. 

(3)The problem of obtaining the training data set before the first use of the framework should be solved.

Things worth being improved in this peper: (Summary)

device that a certain amount of computational power is necessarily needed

Launch the framework.

This can refer to how Name Node&Secondary Name Node cooperate to update FsImage in HDFS.  

AWS Cloud Service

On the right is the architecture of the AWS Greengrass. span style=\"font-size: inherit;\

We need edge computing platform to play the role of moving computing power closer to data source to achieve the effect of reducing latency and saving network bandwidth.

These steps should be taken first before actually use the framework.

device that only need  relevantly lower computational power

Invokes the  Predictor.Predict to obtain the latencies and costs information for edge and cloud lambda functions.

How to design the architecture of the edge severless platform so that it can be deployed in most edge scenarios and provide efficient services?

   This paper propose a framework that dynamically determines where to excute serverless functions so as to optimize developer-specified criteria.

N

Q2: What is the solution to this problem in this paper and what is the key to this solution?

It seems that the only reason for this paper to study this task placement problem in the context of severless computing is that serverless computing is becoming increasingly popular for both cloud and edge platforms.

To be more specific

有黄色背景的内容是我个人的一些思考

Invokes the Executor.

Invokes the Uploader.

Y

Edge lambda function is selected? 

AWS IoT SDK

Edge-cloud collaboration is required to enhance strengths and avoid weaknesses.

Adjust the parameters of the Predictor.

AWS IoT Greengrass Core

This is the text and picture description of edge pipeline in the paper which it's not resonable enough. Here are the two points that confuse me:1. The \"AWS IoT Core service in the cloud\

(4)The latency and cost information about the lambda functions' execution should be collected and added into the training data set to produce an effect while the framework and the application is running. And the update of the prediction models should be done in an on-the-fly way.