======= Review 1 ======= *** Paper Summary: Please summarize the paper in your own words. The paper deals with handovers in Cloud Radio Access Networks. The specific kind of handover addressed by the paper is the one occurring in a virtualized radio access. In this case a user, due to its mobility, can or cannot change the Baseband Processing Unit (BBU) while changing the Remote Antenna (RRH). A careful design of the BBU-RRH mapping can minimize the total number of such kind of handovers. *** Strengths: What are the main reasons to accept the paper? You may comment on the importance of the problems addressed, the novelty of the proposed solutions, the technical depth, and potential impact. Your overall rating should be supported by your review. The BBU-RRH mapping is a timely research topic and matches an active area of research. The study on the effects of this mapping on the handovers that can happen when a user changes the RRH are interesting and novel. The proposed online approach based on time varying network graphs is significant. *** Weaknesses: What are the main reasons NOT to accept the paper? Again, think about the importance of the problems addressed, the novelty of the proposed solutions, the technical depth, and potential impact. Your overall rating should be supported by your review. For the time being the proposed approach is constrained by the capacity of the system while it does not consider limitations that can be due to latency issues. The performance analysis is mainly based on voice calls while the handover issues in the current and future cellular systems may have an important impact on other kind of services (e.g. video ones) that have other capacity and delay constraints. The use of voice calls is too restrictive. The authors do not explain which are the dynamics of these kind of handovers. How frequently a mobile user (urban environments for instance) changes the BBU in a typical cloud based architecture? The complexity of the propose approach could be an issue. No comments on this a re provided. Specifically, an evaluation of the online algorithm complexity in the specific case study is missing. *** Quality of Writing: What is the presentation quality of this paper? A paper not well-written is not good for INFOCOM reputation and will have difficulty in attracting citations. Your overall rating should take this into consideration. The quality of the writing is good. Only some typos: • Pag 2 a virtualized RAN resultresults • We uncover The title of the paper could a little made more specific, for instance by mentioning these BBU handovers! *** Additional Comments: Additional comments (if any) that you would like to provide to the authors. Please do not repeat what you stated above. If none, leave the following blank. How complex is the derivation of the network graph snapshots in case of cloud RANs? Which is the impact of the proposed scheme in terms of signaling cost for moving a user from a BBU to another? *** Overall Rating: Your overall rating (based on strengths, weaknesses, and quality of writing). accept - top 20% of all papers assigned to me for review (4) ======= Review 2 ======= *** Paper Summary: Please summarize the paper in your own words. The paper describes a new type of handover occurring in CRAN architectures which is due to the dynamic reallocation of RRHs to different BBUs at different traffic loads. The paper analyzes this type of handovers and proposes a dynamic RRH reconfiguration algorithm aiming at minimizing the overall number of handovers, both those related to mobility and those caused by the RRH reconfiguration. The performance of the algorithm are tested in realistic call data sets of two African cities. *** Strengths: What are the main reasons to accept the paper? You may comment on the importance of the problems addressed, the novelty of the proposed solutions, the technical depth, and potential impact. Your overall rating should be supported by your review. The paper introduces the new type of handovers caused by RRH configurations in CRAN scenario and highlights its importance. The paper presents an elegant and effective solution approach based on time-varying graphs and a variation of a graph clustering approach. Results show that neglecting the introduced type of handovers can indeed results in more overall handovers. *** Weaknesses: What are the main reasons NOT to accept the paper? Again, think about the importance of the problems addressed, the novelty of the proposed solutions, the technical depth, and potential impact. Your overall rating should be supported by your review. The proposed solution consists of two parts: one dealing with oracle-based scenarios, the other considering future handover predictions. The latter contribution is too simplistic, the way future behavior is predicted can be much more sophisticated than assuming nothing will change. In addition, it is not clear how the number of active BBUs is adapted to demand changes over time. Finally, how to deal with predictions uncertainness has not been properly discussed. The authors claim to have used a realistic data set, however its data have been so heavily manipulated to be adapted that they could be considered equivalent to any other arbitrarily generated random data set. The proposed solution considers a capacity model based on GSM calls, which is definitively outdated in the current mobile data world. *** Quality of Writing: What is the presentation quality of this paper? A paper not well-written is not good for INFOCOM reputation and will have difficulty in attracting citations. Your overall rating should take this into consideration. The paper is generally well written and well organized. The reviewer has only some comment on Figg. 7 and 9: 1) they include a “Resources” curve that is never mentioned in the text, 2) subfigures of Fig.7 are too small to be compared by eye. *** Additional Comments: Additional comments (if any) that you would like to provide to the authors. Please do not repeat what you stated above. If none, leave the following blank. The reviewer has some concerns on the choice to predict the expected user migration in the future exactly at the same level of the current values. This basically destroys any prediction and makes the introduction of time-varying graphs useless. Indeed, this means that the optimization can be simply done on the currently measured network status. The reviewer understands that the future prediction may imply some difficulties, however using the proposed approach wastes past information as well. Indeed, the system could be first dimensioned on past statistics, and then modified online according to the current measurements. The issue of how to consider the unavoidable prediction uncertainness in the proposed algorithm and its effect on its outcome should be better discussed as well. The considered data set consists of hourly samples of the number of calls and their average duration between any pair of BSs of the considered cities. The steps followed to generate user mobility over this data set and to obtain 10-minute snapshots are completely arbitrary. The assumptions made to manipulate the data set are so strong that basically any other random data set could have been considered. Finally, two additional comments are: - The reviewer believes that considering a capacity model based on calls makes the paper old. It is better to consider data-oriented capacity models with some threshold on the minimum perceived throughput. - In the discussion of the results, handovers are always globally considered. It would be much more interesting to see which share is due to the user mobility and which one caused by RRH reconfigurations *** Overall Rating: Your overall rating (based on strengths, weaknesses, and quality of writing). accept - top 20% of all papers assigned to me for review (4) ======= Review 3 ======= *** Paper Summary: Please summarize the paper in your own words. The article sets out to answer the question of how mobility management should be handled in a CRAN architecture. The authors consider an additional type of handover, which happens when software functions for an RRH migrate from one BBU to another, which they call a reconfiguration handover. They model mobility handovers and reconfiguration handovers using a time-varying graph. *** Strengths: What are the main reasons to accept the paper? You may comment on the importance of the problems addressed, the novelty of the proposed solutions, the technical depth, and potential impact. Your overall rating should be supported by your review. The modeling of what the authors refer to as ‘reconfiguration handovers’ is a valuluable aspect of CRAN management. Good positioning of the contribution with respect to the literature. Clear model and solution approach. A well written paper. *** Weaknesses: What are the main reasons NOT to accept the paper? Again, think about the importance of the problems addressed, the novelty of the proposed solutions, the technical depth, and potential impact. Your overall rating should be supported by your review. While the use of real operator data from two African countries for performance assessment is commendable, the data used is old (2012-13), especially by wireless network standards, and is restricted to voice traffic. It also requires the authors to make a number of assumptions about call arrivals and frequency of handover. Given all these limitations, I don’t think the data is solid enough to draw the conclusion that not considering traffic estimation does not have a big impact in the results, as claimed in Section V.B. *** Quality of Writing: What is the presentation quality of this paper? A paper not well-written is not good for INFOCOM reputation and will have difficulty in attracting citations. Your overall rating should take this into consideration. The writing is generally good. One exception: ‘It commends’ (?), p.1. *** Additional Comments: Additional comments (if any) that you would like to provide to the authors. Please do not repeat what you stated above. If none, leave the following blank. The authors use four paragraphs to introduce the concept of CRAN, which is familiar to the immense majority of the INFOCOM public. I think the statement that the online solution performs “better” than the oracle from the point of view of handovers is odd, since this is a result of increased call drops. I recommend that this result, in section VIII, be rephrased. (One could decrease handovers all the way to zero, by increasing call drops.) *** Overall Rating: Your overall rating (based on strengths, weaknesses, and quality of writing). borderline - top 50% of all papers assigned to me for review, but not top 20% (3) *********************************************************************************************************************************************************************************************** Comments for previous version submitted at INFOCOM 2017 *********************************************************************************************************************************************************************************************** ======= Review 1 ======= *** Paper Summary: Please summarize the paper in your own words. The paper studies effective allocation of baseband units (BBUs) to remote radio head (RRHs) in a cloudified RAN setting. The objective is to minimize an accurate measure of handover (due to both mobility and reconfiguration). This done through construction of a graph model and solving its clustering in an online and also a prescient manner. *** Strengths: What are the main reasons to accept the paper? You may comment on the importance of the problems addressed, the novelty of the proposed solutions, the technical depth, and potential impact. Your overall rating should be supported by your review. The paper provides a convincing model of cloudification of the RAN and states assumptions needed and used in an honest manner. Reducing the problem to a graph partitioning problem is also interesting but not surprising. A standard model (modularity) is used and implemented and computational results presented with a good discussion of assumptions given that the data correspond to a 2G network. *** Weaknesses: What are the main reasons NOT to accept the paper? Again, think about the importance of the problems addressed, the novelty of the proposed solutions, the technical depth, and potential impact. Your overall rating should be supported by your review. There are several weaknesses that are explicitly stated on the 2nd column of page 3: o such that an edge et i j 2 Et m exists between it and jt if the cells corresponding to RRHs i and j in the Voronoi diagram of the access network share a common border Voronoi diagram of the access network share a common border. <-- neighboring cells need not be physically neighbors o We remark that, in this work, we do not consider RRH cooperation, enabled for example by using coordinated multipoint (CoMP) techniques <-- this is a natural and practical use of multi BBU to multi RRH association and its absence is a serious limitation o We do not consider this geographical dimension in our work, as it opens new and complicated problems, such as the optimal placement of the data centers hosting the BBUs. <-- There is strict distance limitation between BBU and RRH that has to be enforced for 1-2 m latency to be met Another serious limitation is the computational results. This reviewer could not see a baseline relative to which improvements due to the proposed approach was obtained. *** Quality of Writing: What is the presentation quality of this paper? A paper not well-written is not good for INFOCOM reputation and will have difficulty in attracting citations. Your overall rating should take this into consideration. The paper is well written and generally easy to follow. The exception being description of the graph model which is too terse. *** Additional Comments: Additional comments (if any) that you would like to provide to the authors. Please do not repeat what you stated above. If none, leave the following blank. *** Overall Rating: Your overall rating (based on strengths, weaknesses, and quality of writing). borderline - top 50% of all papers assigned to me for review, but not top 20% (3) ======= Review 2 ======= *** Paper Summary: Please summarize the paper in your own words. The paper considers the problem of handovers in cloud radio access networks (CRAN) where the association between base band units (BBUs) and remote radio heads (RRHs) is dynamically modified. The authors propose an on-line algorithm for BBU-RRH association that is able to minimise the number of handover considering traffic predictions and capacity constraints. *** Strengths: What are the main reasons to accept the paper? You may comment on the importance of the problems addressed, the novelty of the proposed solutions, the technical depth, and potential impact. Your overall rating should be supported by your review. The paper is well written and the approach proposed is technically sound. The model appears accurate enough for capturing the key system features required for describing the effect of dynamic associations BBU-RRHs on handovers. *** Weaknesses: What are the main reasons NOT to accept the paper? Again, think about the importance of the problems addressed, the novelty of the proposed solutions, the technical depth, and potential impact. Your overall rating should be supported by your review. The motivation of the work is extremely weak. The association between users and BBUs follows objectives that can hardly include handovers. The main reason why operators may want to rearrange dynamically association is per-user performance in terms of throughput rather than handovers. Not only, for the optimisation of mobility related targets like handovers, the estimation and prediction of traffic variations is very critical and may affect any possible gain. *** Quality of Writing: What is the presentation quality of this paper? A paper not well-written is not good for INFOCOM reputation and will have difficulty in attracting citations. Your overall rating should take this into consideration. The quality of writing is fairly good. *** Additional Comments: Additional comments (if any) that you would like to provide to the authors. Please do not repeat what you stated above. If none, leave the following blank. *** Overall Rating: Your overall rating (based on strengths, weaknesses, and quality of writing). likely reject - bottom 50% of all papers assigned to me for review, but not bottom 20% (2) ======= Review 3 ======= *** Paper Summary: Please summarize the paper in your own words. This paper proposed a solution which aims to minimize the mobility handovers. In particular, two types of handover operations are jointly considered including mobility handover and reconfiguration handover. The former one is widely observed in the traditional cellular networks which is caused by changing the associations with different BSs as UE moves. The latter one is the new handover caused by different associations of RRH with BBU as time proceeds. Two algorithms are proposed to minimize the total handover costs. The first algorithm assumes the perfect knowledge of the user mobility and traffic patterns. The second one assumes the user and traffic status will be estimated on-the-fly. The performance of the proposed algorithms are verified based on the simulations and traffic traces from Orange GSM Networks *** Strengths: What are the main reasons to accept the paper? You may comment on the importance of the problems addressed, the novelty of the proposed solutions, the technical depth, and potential impact. Your overall rating should be supported by your review. The proposed reconfiguration handover is new, which does cause the new challenges of dynamic RRH clustering for C-RAN. In addition, using the actual data traces from real operators may give some valuable insights into the practical performance of the proposed algorithms *** Weaknesses: What are the main reasons NOT to accept the paper? Again, think about the importance of the problems addressed, the novelty of the proposed solutions, the technical depth, and potential impact. Your overall rating should be supported by your review. First, the paper did not consider the RRH cooperation, which is however one of the key advantages of C-RAN. Through dynamic RRH clustering, we can significantly increase the network capacity and reduce inter-cell interference. As a result, in the formulated optimization problem, the capacity gain of RRH clustering need to be considered in the constraint, instead of assuming the non-cooperation of RRHs. Second, although the authors use the weight to indicate the handover cost, how to analytically or numerically estimate the cost is not clear. Specially, for reconfiguration handover, authors may need to consider the paging and registration cost caused by changing the BBU-RRH associations. Third, for the proposed on-line algorithm, authors assume that that there are not variance in the mobility and traffic patterns as time proceeds. This simplifies the analysis but is not that realistic. *** Quality of Writing: What is the presentation quality of this paper? A paper not well-written is not good for INFOCOM reputation and will have difficulty in attracting citations. Your overall rating should take this into consideration. The writing still has a room for improvement. *** Additional Comments: Additional comments (if any) that you would like to provide to the authors. Please do not repeat what you stated above. If none, leave the following blank. *** Overall Rating: Your overall rating (based on strengths, weaknesses, and quality of writing). likely reject - bottom 50% of all papers assigned to me for review, but not bottom 20% (2) *********************************************************************************************************************************************************************************************** Comments for previous version submitted at MobiCom 2016 *********************************************************************************************************************************************************************************************** =========================================================================== MobiCom'16 Review #359A --------------------------------------------------------------------------- Paper #359: User Mobility in Dynamic Cloud Radio Access Networks --------------------------------------------------------------------------- Overall merit: 2. Weak Reject Reviewer expertise: 3. Knowledgeable ===== Paper summary ===== This paper presents a way to reduce the number of handovers in Cloud Radio Access Networks (CRAN) that consist of baseband processing units (BBUs) and remote radio heads (RRHs). In a virtualized environment of CRAN, the paper reveals a new type of handover called Reconfiguration HandoverTh (RHO) in CRAN where the mapping between BBUs and RRHs can change dynamically even for a static user. Then, the paper presents an analytical model using a time-varying graphical model to analyze such mapping between BBUs and RRHs. Based on the model, authors proposed a modified Louvain method to obtain an optimal solution, assuming all network activities are known. In addition, the paper proposes an online solution that assumes there is only past and short-term prediction of future network activities. Finally, the paper shows evaluation results on their algorithm using Call Detail Records (CDRs) in an area in Ivory Coast. ===== Paper Strengths ===== The paper identifies new problem, called RHO -- reconfiguration-induced handover in CRAN. Authors apply the time-varying graph technique to model the identified problem and solve the problem with a novel method. With large-scale Call Detail Records, authors proves the effectiveness of their algorithms. ===== Paper Weaknesses ===== Please see below: ===== Comments for authors ===== This paper identifies an interesting problem in CRAN, called RHO. This problem would significantly impact user experience and resource utilization. With the real-life call record in large areas, authors show their proposed scheme indeed reduce RHO (by 20%). Although the problem identified is critical and novel (first work in CRAN according to authors), it is hard to estimate the impact of the improvement (i.e., 20%) in real environment. Further, there are several concerns and suggestions on this paper as follows: The paper illustrates a case when this new type of handover (RHO) happens as shown in Figure 2. RHO happens to even a static user when a BBU for a given RRH has changes for some reason. One reason might be the case when a new BBU is created to handle heavier load and forward user information of an old BBU to a new BBU. This scenario is conceptually possible. However, it might be better to see authors' statements to make sure that this problem is artificial or not. Regarding the time-varying graphical model, authors map a node in graph to an RRH and an edge to a handover candidate. Their framework is to find a optimal clustering to reduce handover. To solve the problem, authors have added a constraint checking procedure to the original Louvain method in order to capture the BBU capacity. It is better to see analysis of their algorithms such as complexities. In evaluation, the model for BBU capacity is missing. Authors claim in sec. 7.1 that their algorithm induces 23% or 24% less handovers compared to traditional RAN. However, it would be better if the paper presents the results in graph form. Further, its comparison with other areas would be helpful. =========================================================================== MobiCom'16 Review #359B --------------------------------------------------------------------------- Paper #359: User Mobility in Dynamic Cloud Radio Access Networks --------------------------------------------------------------------------- Overall merit: 3. Accept Reviewer expertise: 4. Expert ===== Paper summary ===== The authors addressed the association of BBUs and RRHs in CRAN and put the optimization in the context of mobility management. The aim to reduce the number of handoffs (but this handoff definition is different from the conventional one) and thus proposed an optimization over time-varying graph. They used real traces and evaluated the proposed solutions. ===== Paper Strengths ===== + Well written and well-executed ===== Paper Weaknesses ===== - The problem of reducing the number of changes of BBU-RRH associations (or called as handoff) might not be that critical in CRAN. ===== Comments for authors ===== This is a nice read. The paper is well written, organized and executed. The handoff concept in the paper is totally different from the one used in today cellular networks. In fact, the authors focus on the switch of BBU-RRH associations. This is useful to reduce the migration overhead but user mobility in the title is misleading. The following scenarios requires mobility support but it seems out of the authors’ interests. The user moves out of one RRH and into a new RRH which is associated with the same BBU, it will not trigger the change of association but it still requires handoff between RRHs. This case is not considered in the paper. I am skeptical in motivation of this work. How much gain can CRAN obtain if it reduces the number of BBU-RRH associations. The paper seems to work under the premise that VM migration (BBU migration) in CRAN is very costly or time-consuming or critical to user experience. However, this is not justified. Moreover, it is not clear what 22% reduction means? What might be benefits to user experience or system efficiency or something like. =========================================================================== MobiCom'16 Review #359C --------------------------------------------------------------------------- Paper #359: User Mobility in Dynamic Cloud Radio Access Networks --------------------------------------------------------------------------- Overall merit: 1. Reject Reviewer expertise: 3. Knowledgeable ===== Paper summary ===== While Cloud Radio Access Networks (CRANs) have drawn much interest in recent years, the authors argue that existing work has overlooked the mobility aspect and highlight a type of handover specific to CRANs, i.e., when the remote antenna becomes associated with a different baseband processing unit. The authors then propose two algorithms, one offline, “oracle”-style, and the other online, to optimize the mapping between antennas and processing units to reduce the number of handovers. Both algorithms are evaluated, presumably with simulations, using synthesized scenarios based on call record dataset from Abijian, Ivory Coast. ===== Paper Strengths ===== The paper explicitly considers mobility and handover in a CRAN setting. The paper tried to use a real data set. ===== Paper Weaknesses ===== It is not clear what problem the paper is really addressing. Most of the discussions are very hypothetical. Both algorithms, whether the oracle version or the online version, assume global knowledge of mobility and traffic patterns, which are not at all practical. The evaluation is very weak. The wording regarding the Abijian dataset in the abstract/intro is very misleading, since, in the end, almost everything was synthesized based on typical stochastic models rather than anything real. Also, I can only infer that some simulations were run. ===== Comments for authors ===== I agree mobility management is an important part of the future mobile network architecture, and it certainly is helpful to discuss that in the context of CRANs. However, I’m not sure what to take away from the paper. — I suppose the paper tried to discuss handover strategies, but these depend on both the CRAN design and the mobility characteristics. For the latter, we might draw on experience in existing mobile networking scenarios, even though there are still issues of wide area vs local area. For the CRAN design, I don’t think there is even a consensus on a “standard” design yet, and this is completely glossed over in the paper, as if there was some reference architecture already. With these in mind, I doubt it is the right time to discuss handover strategies at this stage. — Even if we want to consider handover strategies now, these should be considered jointly with the CRAN design. For example, some recent systems (e.g., Argos and BigStation) basically see massive MIMO as one realization of the CRAN vision. In that setting, user handover is automatically taken care of by always leveraging the antennas within range of the user device. The notion of “association” is weakened, and “handover” in the conventional sense would be fairly seamless. The challenge then shifts to keeping track of the right remote antennas and the corresponding channel information. Other CRAN designs would impose different constraints. — Taking a step back and assuming we are following the implicit CRAN design in Figure 2, I’m then confused what the algorithms are trying to show. Sure, it is helpful to minimize the number of handovers, and the oracle algorithm shows the bound. Now is that meant to make a case for CRAN or just showing the best performance of a non-oracle algorithm? The evaluation in the paper follows the former train of thought, but the comparison basis is not clear. In any case, previous papers have made cases for CRAN, so I’m not sure what the paper is adding. Regarding the online algorithm, I’m also unclear what it is intended to show. It still assumes global knowledge of the mobility and traffic patterns, which is hardly realistic. — The evaluation is weak. I’m guessing that some simulations were run, but there are no details at all, not to say simulations are usually not as useful as prototype implementations. Anyway, trace-driven simulations can still be useful, although I wonder why the authors thought a dataset from Abijian almost 5 years ago would be representative. But even disregarding that, the actual simulations still seem disappointing. In the abstract/introduction, the authors give the impression of replaying a real data set, but in practice, almost anything critical (call arrival statistics, call duration, mobility) was synthesized. This is hardly “using a real data set”! =========================================================================== MobiCom'16 Review #359D --------------------------------------------------------------------------- Paper #359: User Mobility in Dynamic Cloud Radio Access Networks --------------------------------------------------------------------------- Overall merit: 2. Weak Reject Reviewer expertise: 3. Knowledgeable ===== Paper summary ===== This paper presents solutions based on time-varying graph representations for handling two types of mobility in cloud-based radio access networks - the traditional UE mobility and the Remote Radio Head (RRH) mobility due to RRHs being assigned different baseband processing units (BBUs). A data set from Ivory Coast is used to evaluate the proposed solutions. ===== Paper Strengths ===== The proposed solution can possibly reduce mobile handovers. ===== Paper Weaknesses ===== The problem of RRH handover seems contrived. The evaluation using the GSM dataset appears to be very artificial and not convincing. The authors do not seem to be aware of some of very relevant related work. ===== Comments for authors ===== The RRH handover problem seems contrived. The authors have provided no evidence that such a problem exists. Moreover, even if we assume that this problem could possibly exist in a CRAN, why cannot the handover of an RRH to another BBU transfer the old BBU's air resources to the new BBU for existing connections? Why would a fancy over the air UE handover be needed? The evaluation uses a GSM voice dataset and makes several assumptions (including mobility assumptions because that information is not included in the data) to somehow show that the approach proposed in the paper reduces the handovers. The traditional UE handover for CRANs has been very nicely addressed in the FluidNet paper by Sundaresan et al [IEEE/ACM ToN 2015]. So the authors are not really solving a new problem here. They do not refer to this work. Another paper the authors might want to look at is on Scaling the LTE Control Plane by Sundaresan et al [CoNext 2015] for understanding load balancing and state migration issues. =========================================================================== MobiCom'16 Review #359E --------------------------------------------------------------------------- Paper #359: User Mobility in Dynamic Cloud Radio Access Networks --------------------------------------------------------------------------- Overall merit: 1. Reject Reviewer expertise: 3. Knowledgeable ===== Paper summary ===== This paper proposes algorithms to improve the mapping between BBUs and RRHs in the CRAN context. The goal is to reduce the number of handovers in the system. ===== Paper Strengths ===== Mobility management in the cloud context is an important topic for future cellular networks. The paper tries to formulate one of the problems. ===== Paper Weaknesses ===== The graph theoretic formulation for optimal mapping between BBUs and RRUs is straightforward, and online algorithm lacks performance analysis. ===== Comments for authors ===== It is an algorithm paper, but the proposed algorithm is pretty straightforward and lacks novelty, as well as rigorous performance analysis (e.g., competitive ratio for the online algorithm). Several issues are missing: 1. How do you deal with failing BBU/RRU, which runs as VMs in the cloud? 2. How significant is the negative impact due to the increased handover stemming from improper BBU and RRU mappings? There is no empirical or analytical results on it. 3. The evaluation has been using very simplistic assumptions. It does not assess or discuss how to support data sessions. The call model is also simplistic. Consequently, the evaluation results are suspicious.