Date: Thu, 16 May 2002 23:50:11 -0500 (CDT) From: Yu Hu YU HU An Empirical Study of Epidemic Algorithm in Large Scale Multihop Wireless Networks [Deepak Ganesan, Bhaskar krishnamachari et.al.] Main contributeion of the paper: This paper is a detailed analysis on epidemic algorithm in large scale multihopwireless network. Even pretty simple epidemic algorithm will show a complicatedview. We can find some phenomeno from this paper and authors explained them to us based on experiments of separate layers. I think the main contribution of this paper is that it showed us a detailed analysis on epidemic algorithm and itsmethod by use of struture of layers. Critique the main contribution: Significance : 4( significant contribution) Methodology: In this paper, authors used two set of experiments for two different purposes. Also authors did analysis basing on different layers. I think these are good methods because we can tranfer complex problems into some simple andclearer subproblems. And we can easily analyze and find reason. The most important limitation of the approach: The size of experiment is not huge. And I think maybe some problems about scalalibity can't be correctly analyzed by use of these experiments. Most interesting ideas: Every layer of protocol has its contribution to performance of epidemic algorithm on wireless network. This is really interesting because we can make use of it to optimize our design of every layer of protocol and can easily control sometrade-off by modifing some protocol layers. Weakness&Questions: Analyzing problems on different layer is a good idea. However, I want to ask whether some important reasons that are from combinaiton of some layers are neglected ? Maybe it will impact the result much . Comments: This is a good paper, I like it. From it, I got a clear view of epidemic algorithm on wireless network and I learn how to analyze issues based on different layers. Performance Comparison of Two On-Demand Routing Protocols for Ad Hoc Networks [Charles E.Perkins, Elizabeth M.Royer et.al.] Main contributeion of the paper: This paper main job is comparing the performance of DSR and AODV. It helped us to learn these two dominent on-demand routing protocols fro ad hoc network, their performance and performance difference. Critique the main contribution: Significance : 3(modest contribution) Methodology: I think the method used by authors here is pretty straitforward.Implementing these protocol in simulation model is a normal means. Most interesting ideas: In this paper, we found that interplay between the routing and MAC layers couldaffect performance significantly, which is worthy of being considering. And we can remember the former paper, maybe it will challenge its separating layer method. Weakness&Questions: Maybe authors should introduce more information about why these two protocols become dominent and these performance differences are because of what charateristics of the two protocols? Comments: I think this is a normal paper. Anyway, we can learn the performances and difference of two important on-demand protocols, which will help us to learn about the ad hoc network . ################################### ################################### Date: Fri, 17 May 2002 00:02:34 -0500 (CDT) From: Xinghua Shi Reading10: Paper1: Performance Comparison of Two On-demand Routing Protocols for Ad Hoc Networks Contribution: The paper presents a systematic performance study of two dynamic routing protocols for ad hoc networks,i.e. Dynamic Source Routing protocol(DSR) and the Ad Hoc On-Demand Distance Vector protocol(AODV). It presents some useful analysis based on results from simulation of the two protocols in different scenarios. This research has a thorough study of the two protocols and provides useful information for choosing protocol in implementation. Rate: a. 3 in significance: This paper has moderate contribution because it provides a detailed studies and comparison of the two protocols. But it just presents the comparison and not too many new ideas presented. b. 4 in convincing of methodology: The argument is well organized. It first introduces the two protocols to be discussed. Then it builds simulation and gets results. The following analysis is based on these results and detailed explained. The strongest and/or most interesting ideas in the paper: a. Neither protocol has any mechanisn for load balancing,that is, choosing routes in such a way that the data traffic can be more evenly distributed in the network. b. DSR always demonstrates a lower routing load than AODV. c. MAC load is a good measure for predicting application performance. d. Mobility affects the performance of the two protocols differently. The most striking weaknesses in the paper: a. Simulation model is not well explained and founded. b. Only statistics and analysis based on the simulation. Lack insight into their effect on the real implementations. Questions: a.What's the analysis and comparison deal with the practical use of these two protocols? b.Are there other protocols in this field? Which one is used widely in practice? Paper2: An Empirical Study of Epidemic Algorithms in Large Scale Multihop Wireless Networks Contribution: This paper presents detailed empirical data from studies of relatively large scale wireless network configurations to serve as a basis for algorithm design in large scale multihop wireless networks. The authors discuss the experiments results from an experimental platform. In analyzing the contributions of various layers, they define and study useful metrics at each level. According to this, they point out some implications for the design of protocols for large scale wireless networks. They also propose some open questions for discussion with further research. Disadvantages: The results and analysis are only based on their experiments, so they maybe not exactly right in a general sense, though the implications are useful. Rate: a. 3 in significance: This paper has moderate contribution because it has a detailed analysis on their experimental data and presents some implications for algorithm design in this field. But they are only focused on limited experiments. b. 3 in convincing of methodology: The analysis and conclusion are based on the dataset they got. But the argument is merely based on experiental data and results, which lacks general meaning. The strongest and/or most interesting ideas in the paper: a. The aggregate effect of the small differences in the radios and hardware, and slight differences in energy levels of the nodes contribute signicantly to link asymmetries in this regime. b. Even a simple epidemic protocol, flooding, can exhibit surprising complexity at scale. c. Algorithm designs should use a probabilistic abstraction to model connection. The most striking weaknesses in the paper: a. Their experiment set has some restrictions they set to proceed the experiments which maybe biased. Questions: a.Are there some algorithms in use? Do these algorithms follow the implications this paper presents? ################################### ################################### Date: Fri, 17 May 2002 00:09:13 -0500 (CDT) From: Ivona Bezakova Performance Comparison of Two On-Demand Routing Protocols for Ad Hoc Networks [PRDM] 1. State the main contribution of the paper: Simulations of the most recent versions of routing protocols DSR (Dynamic Source Routing) and AODV (Ad Hoc On-Demand Distance Vector) are presented. Four parameters are compared: packet delivery fraction, average end-to-end delay, normalized routing and MAC loads. Some improvements are suggested: (i) using more realistic metric as opposed to number of hops, and (ii) discarding old packets. 2. Critique the main contribution. a. Rate the significance: 2. From the previous work section I concluded that most of the results were known, they were just confirmed under a slightly modified setting (new versions of the protocols). b. Rate how convincing: 2-3. This depends mainly on my trust in the simulator, which is not described at all. (I could follow the reference link but I think that at least an overview of the simulator and the probabilistic model of changes should have been presented.) 3. What are the three strongest and/or most interesting ideas in the paper? Unfortunately, I did not find any. 4. What are the three most striking weaknesses in the paper? - As mentioned above, intuitive description of simulator is missing. - What is random waypoint method and why is this one chosen for the simulations? --------------------------------------------------------------------------- An Empirical Study of Epidemic Algorithms in Large Scale Multihop Wireless Networks [GKWCEW] 1. State the main contribution of the paper: A simulation of behavior of a static wireless network using real devices spread over a reasonably big area. The authors flood the network (run BFS) and observe various parameters of the network, depending on the radio adjustment settings. The parameters include packet loss, connectivity radius, asymmetry of links, latencies, etc. Some of the conclusions are: (i) 95% of nodes are covered within 2 seconds, while the rest 5% takes another second (ii) 5-15% of the links are asymmetric. 2. Critique the main contribution. a. Rate the significance: 4. I find a study using real devices and their actual features and limitations very interesting. b. Rate how convincing: 4. Well-written, easy-to-read paper - the results are stated clearly. c. What is the most important limitation of the approach? The nodes do not move. 3. What are the three strongest and/or most interesting ideas in the paper? - Using actual devices and not a computer-based simulation. 4. What are the three most striking weaknesses in the paper? - I did not find any striking weaknesses, rather some suggestions for future work - see 6. 5. Name three questions that you would like to ask the authors? - How much did the experiment cost? 6. Detail an interesting extension to the work not mentioned in the future work section. - Compare this study with computer-based simulations and evaluate their performance. - How to design a non-static experiment using real devices? - What is the influence of the radio adjustment settings on the battery life? - Design a corresponding physical model, perhaps some of the conclusions could be derived theoretically. - Other topology/algorithms, can we predict behavior using these statistics and based on that design a good protocol (for some problem)? ################################### ################################### Date: Fri, 17 May 2002 04:58:32 -0500 From: Oleg Pashko Performance Comparison of Two On-Demand Routing Protocols for Ad Hoc Networks 1. Performance of 2 on-demand routing protocols for mobile ad hoc networks: Dynamic Source Routing (DSR) and Ad Hoc On-Demand Distance Routing (AODV) is compared by extensive simulations. It has been established that the differences in the protocols lead to substantial performance differences. The differences were studied in the context of varying network load, mobility & network size. 2a.2.5 Useful observation about behavior of 2 existing protocols and routing schemes. 2b.Simulations seem to be well designed, but the amount of empirical data exceeds the data analysis by a great deal. 2c.I feel that there was a lot of data presented, but not enough evaluation and analysis. 3.Observation on how to improve DSR performance (by mechanisms to expire routes and/or determine freshness of routes in the route cache). 4.same as 2c. 5-7. How general are the conclusions? Do the protocols for mobile ad hoc networks have similar behavior? An Empirical Study of Epidemic Algorithms in Large Scale Multihop Wireless Network 1.A large scale empirical study of the epidemic algorithms at various layers of the protocol stack. 2a.4. extensive and interesting research of an important class of algorithms. 2.b. Well designed large scale experiments in real multi layered network systems. 2.c.Some results are trivial & quite predictable (like redundancy of backward links, long links straggling nodes are to be expected). 3.It was very interesting for me to find out about the behavior of the epidemic messages propagation in various layers (physical, link, medium, network,application, etc).Interesting data about useless broadcasts, collision at the MAC layer,etc.Particularly interesting, since these not simulations, but large scale measurements. 4.same as 2c. 5-7.How to design algorithms with probabilistic abstraction to the model?