Ip Network Coordinate System Stability Enhancement Technology

As the scale of internet grows exponentially, the architecture of internet has also become more and more complicated. Distributed internet services, such as P2P service, take more and more proportion in internet services. These services can get benefit from the improvement of the route algorithm as well as the accurate neighbor selection. Round trip time, bandwidth and so on are useful metrics to evaluate the performance of internet. Especially Round trip time is one of the most importance metrics which is usually called network distance. IP network coordinate system, which makes use of RTT between parts of host in the network, constructs a virtual geometric space with assigning each host one coordinate. The geometrical distance estimates the real distance in the internet space. However traditional IP coordinate system is vulnerable in face of malicious nodes and distance prediction lack accuracy and stability when using raw RTT samples. This paper focuses on improving the stability of IP coordinate system and the contribution includes two parts:1. Traditional network coordinate system, which lack security mechanism, is suffering from vulnerability in presence of attackers, which results in low prediction accuracy as well as wrong neighbor selection. To solve the vulnerability of the coordinate system, this paper firstly presents a landmark-based coordinate system with defending against malicious behaviors (LCSD). LCSD adopts the trust mechanism based on the recommendation from both surveyors and reference nodes, to evaluate the credibility of node. Surveyors and reference nodes observe each ordinary node, obtains experience to evaluate the trust. Nodes with high reputation can update the coordinate and be chosen as reliable neighbor. Attackers, which have low reputation, are forbidden to behave, so the malicious behaviors can be defended. The final simulated experiments verify that in presence of the malicious nodes, LCSD can predict the distance more accurately, choose the neighbor hosts more reliably, and system is more reliable.2. Moreover, in IP coordinate system, the coordinate space construction and coordinate update are based on the part of direct RTT measurement between nodes. However in the reality, the RTT measurements between two nodes are not stable all the time, instead, these measurements fluctuate because of queuing or other reasons. This kind of fluctuation also introduces some inaccuracy and instability for the system. This paper proposes a self-adaptive moving filter based on RTT gradient (SMF-RTTG) to pre-treat the input of the coordinate system. The output of the filter is the iterative convolution of the RTT samples and the weight number correspond. This kind of filter can eliminate the random RTT noise to enhance the prediction accuracy of the system. The simulation in Vivaldi system shows that SMF-RTTG can filter out the random RTT noise effectively so as to make the distance prediction more accurate and more stable.