دانلود رایگان مقاله پیش بینی توان مورد انتظار TCP با استفاده از الگوریتم ژنتیک

Abstract

Predicting the expected throughput of TCP is important for several aspects such as e.g. determining handover criteria for future multihomed mobile nodes or determining the expected throughput of a given MPTCP subflow for load-balancing reasons. However, this is challenging due to time varying behavior of the underlying network characteristics. In this paper, we present a genetic-algorithm-based prediction model for estimating TCP throughput values. Our approach tries to find the best matching combination of mathematical functions that approximate a given time series that accounts for the TCP throughput samples using genetic algorithm. Based on collected historical datapoints about measured TCP throughput samples, our algorithm estimates expected throughput over time. We evaluate the quality of the prediction using different selection and diversity strategies for creating new chromosomes. Also, we explore the use of different fitness functions in order to evaluate the goodness of a chromosome. The goal is to show how different tuning on the genetic algorithm may have an impact on the prediction. Using extensive simulations over several TCP throughput traces, we find that the genetic algorithm successfully finds reasonable matching mathematical functions that allow to describe the TCP sampled throughput values with good fidelity. We also explore the effectiveness of predicting time series throughput samples for a given prediction horizon and estimate the prediction error and confidence.

5. Conclusion

The difficulty of predicting the TCP throughput in interference prone WiFi environments is challenging because of different unpredictable effects such as interference, multipath or other users traffic leading to collisions and unpredictable available capacity. In this paper, we propose to model measured TCP throughput samples as a time series and apply the meta heuristic genetic algorithm to match a set of mathematical functions to best represent the time series. By using the set of functions one can predict future samples, given the GA is trained properly. Using our strategy, one can effectively predict TCP throughput evolution over time by just looking at measured throughput samples without the need to have information available from the TCP stack such as estimates on e.g. round-trip-time or packet loss. We have evaluated the impact of different fitness functions and selection algorithms on the accuracy of predictions. When a more accurate prediction is needed, different retraining schemes can be applied at the expense of more computational power required to find the best set of matching functions. Finally, we have demonstrated that the use of feedback strategies always increases the accuracy of the prediction. In order to improve the accuracy even more, a good strategy has to be found that determines when retraining should be applied. As a future work, we intend to develop heuristics that guide when a retraining should be executed, for example based on knowledge about the TCP congestion control phase. Also, we want to study the impact of different sampling intervals on prediction quality as well as study more scenarios such as different interference situations, different bottleneck links, etc. and their impact on the quality of the prediction.