======= Review 1 ======= *** Paper summary: Paper summary The authors propose an approach to solve the assignment problem of base stations to MEC facilities based on clustering and time aggregation. The approach is evaluated by simulating using real mobile datasets of Facebook usage. *** Strengths: What are the major reasons to accept the paper? [Be brief.] + The context of this work (mobile edge computing) is a hot research field + Well written paper with good organization of explanations and experiments *** Weaknesses: What are the most important reasons NOT to accept the paper? [Be brief.] - The main contribution of the paper is the orchestration optimization model in section III.B. The data analysis comes from [37]. I would expect a larger contribution from a full paper, especially in the evaluation section (see comments below). *** Detailed Comments: Please provide detailed comments that will be helpful to the TPC for assessing the paper. Also provide feedback to the authors. The weakness of the paper is the experimental evaluation. In a real setup, I would assume the costs depend on the available bandwidth between BS and MEC resp. two MECs and other aspects, i.e. the amount of data to be moved, the number of switches performed at the same time, etc. The cost functions used in this paper are rather abstract and it is not clear how realistic they are and, therefore, whether the approach proposed by the authors is useful in practice. Same for the number of facilities. In addition, I think it is important that you describe in section V.A how the "combination of Deep Packet Inspection (DPI) and proprietary fingerprinting tools was employed" to get the needed information. The algorithms should be documented since they are part of the work presented in this paper. *** Familiarity: Rate your familiarity with the topic of the paper. Some knowledge (2) *** Contributions: Rate the paper based on the degree of novelty, creativity, impact, and technical depth in the paper. Good (3) *** Presentation: Rate the readability / presentation of the paper. Very good (4) *** Recommendation: Your overall rating. Likely accept (top 20% but not top 10%, significant contribution) (4) ======= Review 2 ======= *** Paper summary: Paper summary The paper focuses on a dynamic, time-varying allocation of Base Stations (BSs) to mobile edge computing (MEC) facilities, so that the users associated with the BSs are allocated enough resource to receive the service from the MEC. The mapping is done by formulating an integer program and taking the usual step of relaxing the integer constraints into linear constraints, solving with CPLEX and rounding the solution back to the integer framework. The paper presents an analysis using actual data from a large operator in major cities. *** Strengths: What are the major reasons to accept the paper? [Be brief.] The paper is cleanly presented and focuses on an interest topic and an active research area (MEC, network virtualization, NFV, etc). The paper uses actual data set from an operator that offers interesting insight into the actual pain points of the operator. The paper's analysis of the data is rigorous and offers some valuable take aways regarding the clustering of the users. The paper's solution is technically correct. *** Weaknesses: What are the most important reasons NOT to accept the paper? [Be brief.] This reviewer feels the problem should be better motivated. See the detailed comments below. The quick take is that the problem of dynamically mapping BSs to MECs needs to be convincingly argued. It seems artificial, since network topology constraints and provisioning issues could lead to static assignment. Also, the authors should demonstrate that MEC capacity is an issue, as I would assume the limiting factor in such a network is not the computing power but the bandwidth (i.e. there is a strong incentive to offload the traffic at the nearest MEC even at the cost of adding resources there). *** Detailed Comments: Please provide detailed comments that will be helpful to the TPC for assessing the paper. Also provide feedback to the authors. The problem of BS to MEC mapping seems to be reasonably optimized with a static mapping. The number of MEC facility is small and designed to capture a set of BSs (and a corresponding set of users). The motivation of the problem needs to be better explained (say, with citation to the literature) to convince it is not artificial. The problem of page 5 is an integer program and is usually hard to solve. This is not discussed in the description of the program. It claims that subgradient optimization can be used to solve the problem, but that assumes first a relaxation of the integer constraint that is not clearly discussed. (It states that "BSs might be fractionally assigned to more than a singe MEC for a time slot", then that a "selective rounding" is performed, but this is an important step that has to be highlighted. It is interesting to mention that facebook is 20% of the traffic on a major operator network in both uplink and downlink. Regarding the evaluation: the methodology of placing the facility with respect to the distance of the BS is a good idea, but it would be better to use the underlying topology. For instance, I would assume the BSs are connected through some fiber loops, and that would be a natural point to connect the cloudlets. This goes back to the practicality session. THe MEC will be placed in the network, and cannot be connected to any BS that is within a certain radius. If the backhaul does not connect the MEC to the BS, then this BS-MEC assignment is impossible. Editorial comments: "leveraging on an existing" -> leveraging an existing.. "whose aims are partially overlapping" "variate" -> "variated"? In this context, "wide"? *** Familiarity: Rate your familiarity with the topic of the paper. Familiar (3) *** Contributions: Rate the paper based on the degree of novelty, creativity, impact, and technical depth in the paper. Good (3) *** Presentation: Rate the readability / presentation of the paper. Very good (4) *** Recommendation: Your overall rating. Accept if room (top 30% but not top 20%, borderline for Networking) (3) ======= Review 3 ======= *** Paper summary: Paper summary This paper presents an orchestration framework for mobile edge computing (MEC) for taking orchestration decisions on base stations to MEC facility assignments. *** Strengths: What are the major reasons to accept the paper? [Be brief.] This paper uses analytics as an input to a model to orchestrate the MEC that utilizes virtual machines to scale up or down computing resources. The simulation results show their solution outperforms the baseline. *** Weaknesses: What are the most important reasons NOT to accept the paper? [Be brief.] Their solution has not considered the management of the control traffic and monitoring traffic. *** Detailed Comments: Please provide detailed comments that will be helpful to the TPC for assessing the paper. Also provide feedback to the authors. This paper proposes a framework for assigning base station and MEC facilities based on analytics. In their model they use 4 weeks of data as the training set. Some issues here are not considered: - how does the model react to sudden change of traffic pattern? E.g., new applications appears that increases the traffic to a certain remote cloud or an application fails in a remote cloud that the traffic volume drops suddenly. - If the remote cloud changes its way to distribute traffic load, does it affect the model? *** Familiarity: Rate your familiarity with the topic of the paper. Some knowledge (2) *** Contributions: Rate the paper based on the degree of novelty, creativity, impact, and technical depth in the paper. Good (3) *** Presentation: Rate the readability / presentation of the paper. Good (3) *** Recommendation: Your overall rating. Accept if room (top 30% but not top 20%, borderline for Networking) (3) ======= Review 4 ======= *** Paper summary: Paper summary The paper presents an analytical framework for orchestrating assignment and migration decisions in Mobile Edge Computing. In particular the authors demonstrate the use of said framework for Base Station to Mobile Edge Cloud association to optimize latency and migration costs. *** Strengths: What are the major reasons to accept the paper? [Be brief.] The paper presents a tangible and novel contribution to the field of resource allocation in Mobile Edge Computing. It is methodologically sound and presents results produced from real-world data. *** Weaknesses: What are the most important reasons NOT to accept the paper? [Be brief.] Glossing over some definitions (notably, computational capacity of the MEC nodes) and not considering some relevant scenarios (changing weights for user and network related costs). *** Detailed Comments: Please provide detailed comments that will be helpful to the TPC for assessing the paper. Also provide feedback to the authors. Section III B is confusing with the great amount of variables present in the text. Some work should probably be done here to make these variables more clear to the reader. Some definition of the computational capacity of the MEC nodes would be desired. Finally, this reviewer would have like to have seen some comparison of the proposed framework using different weights for the network- and user-related costs. Assuming these are always equal is not realistic. *** Familiarity: Rate your familiarity with the topic of the paper. Familiar (3) *** Contributions: Rate the paper based on the degree of novelty, creativity, impact, and technical depth in the paper. Very good (4) *** Presentation: Rate the readability / presentation of the paper. Very good (4) *** Recommendation: Your overall rating. Likely accept (top 20% but not top 10%, significant contribution) (4)