As you read the paper, consider the following questions:
Question 1: What are the two main optimization goals of in-network operator placement? Why are these goals at odds with each other?
Question 2: How does the spring relaxation algorithm help achieve the above goals?
Question 3: The authors propose to perform the placement optimization in two phases: an initial global optimization followed by dynamic local optimizations. What are the benefits and limitations of this approach?