Guest Editorial
|
| Title | Guest Editorial |
| Authors | |
| Abstract | Welcome to this special issue of Journal of Networks (JNW) on “Performance Evaluation of Computer Networks.” The papers included in this special issue are based mainly on selected extended versions of best papers accepted in the 2008 International Symposium on Performance Evaluation of Computer and Telecommunication Systems, SPECTS 2008. This means that these accepted papers have undergone a thorough and a critical review process by experts in the field. Angrishi and Killat explored the invariance condition of the effective bandwidth for finite number of flows traversing a network node. They used the results from large deviation theory and statistical network calculus to determine the minimum number of flows required to observe approximate invariance of the traversing flow’s effective bandwidth. The advantage of the knowledge of effective threshold is that, a simple requirement for a minimum number of independent multiplexing flows at each network node can dramatically simplify the analysis of networks of nodes with different classes of traffic. They proposed two different approaches to determine the effective threshold. The first approach is based on the idea that the probability of the queue being non-empty tends to 0 under many sources limiting condition. The second approach uses the upper bound on the effective bandwidth of the departure flows to determine the effective threshold. Kittiperachol, Sun and Cruickshank presented how real-world TCP can be evaluated using a network simulator. The authors have shown that by combining real-world Linux TCP and INET simulation framework, true dynamics of real-world TCP can be accurately captured in varieties of simulated network environments without having to rely on costly and complicated network experiment. In addition, the applications of LinuxTCP are demonstrated by reviewing congestion control algorithms and displaying congestion window dynamics for each of TCP variants, i.e. bic, cubic, hamilton, highspeed, hybla, scalable, vegas, veno, westwood and compound. Cremonesi, Turrin and Alexandrov designed hierarchical, queuing network performance models that are able to accurately analyze grid architectures and applications. Based on the obtained results, the authors introduced a new allocation policy based on a combination between task partitioning and task replication. The models were used to study two real applications and to evaluate the performance benefits obtained with allocation policies based on task replication. The models have proved to effectively describe the side-effects of parallelization on the overall performance due to resource heterogeneity and node synchronization.Zhani and Elbiaze presented an analysis of prediction performance of training-based models using extensive sets of real Internet measurements. They showed that enlarging training data set does not really improve traffic predictability. They found out that with only 10% of the measurements are quite sufficient to obtain the same prediction error. It was shown that the training-based model can capture, during the training phase, the strong correlation of the traffic with longrange dependence. Then, one lag is practically sufficient to perform quite accurate prediction regardless of the used granularity. Moreover, the authors investigated using exogenous variables as inputs for the α SNF model in order to predict the throughput. When the throughput history could not be available, the use of exogenous variables is very interesting especially when those parameters are easier to measure. The considered exogenous variables are number of packets (pps), the sampled traffic based on packets size expressed in pps or Mbps. It was found that traffic behavior depends on large packets (size≥ 800bytes). Therefore, the performance of throughput prediction is improved when using the throughput of large packets expressed in pps or Mbps as input for the prediction model.Rango, Veltri, Fazio and Marano proposed a new routing protocol for vehicular ad hoc networks. This protocol presents good scalability properties, which is a critical issue for VANET, where a high nodal density needs to be supported and high number of nodes can be involved (>100). This scalability target was obtained through the trajectory concept applied in the forwarding scheme. Finally, Bhatti, and Bateman have shown the measurement of fairness in throughput performance of DCCP Congestion Control ID 2 (CCID2) relative to TCP NewReno, and variants Binary Increase Congestion control (BIC), CUBIC and Compound, all in “out-of-the box” configurations. The used a testbed and end-to-end measurements to assess overall throughput, and also to assess fairness. It is found that in the testbed they used, DCCP CCID2 showed good fairness with NewReno, while BIC, CUBIC and Compound showed unfairness above round-trip times of 25ms. The paper introduced a new metric with which to assess the dynamics of inter-protocol fairness with respect to throughput, a measurement-based approach which allows the new metric to be used easily, a measurement-based, experimental pair-wise assessment of the fairness of NewReno, BIC, CUBIC and Compound TCP against DCCP operating with a TCP-like congestion control (CCID2).We would like to thank all the authors and reviewers for their contributions and devoted efforts. A special thank goes the technical program committee of SPECTS 2008 for their valuable time and superior work. Thanks are also due to the editors and editorial assistants for their fine support. I hope that the work reported here will be useful to your resercah endeavors. |
| Publisher | ACADEMY PUBLISHER |
| Date | 2009-11-01 |
| Source | Journal of Networks Vol 4, No 9 (2009): Special Issue: Performance Evaluation of Computer Networks |
| Rights | Copyright © ACADEMY PUBLISHER - All Rights Reserved.To request permission, please check out URL: http://www.academypublisher.com/copyrightpermission.html. |