Peer-to-Peer (P2P) applications have witnessed an increasing popularity in recent years, which brings new challenges to network management and traffic engineering (TE). As basic input information, P2P traffic matrices are of significant importance for TE. Because of the excessively high cost of direct measurement, many studies aim to model and estimate general traffic matrices, but few focus on P2P traffic matrices. In this paper, we propose a model to estimate P2P traffic matrices in operational networks. Important factors are considered, including the number of peers, the localization ratio of P2P traffic, and the network distance. Here, the distance can be measured with AS hop counts or geographic distance. To validate our model, we evaluate its performance using traffic traces collected from both the real P2P video-on-demand (VoD) and file-sharing applications. Evaluation results show that the proposed model outperforms the other two typical models for the estimation of the general traffic matrices in several metrics, including spatial and temporal estimation errors, stability in the cases of oscillating and dynamic flows, and estimation bias. To the best of our knowledge, this is the first research on P2P traffic matrices estimation. P2P traffic matrices, derived from the model, can be applied to P2P traffic optimization and other TE fields.

A fundamental task of sponsored search is how to find the best match between web search queries and textual advertisements. To address this problem, we explicitly characterize the criteria for an advertisement to be a 'good match' to a query from two aspects (it should be relevant with the query from information perspective, and it should be able to capture and satisfy the commercial intent in the query). Correspondingly, we introduce in this paper a mixture language model of two parts: a commercial model which characterizes language bias of commercial intent leveraging on users' clicks on advertisements, and an informational model which is a traditional language model with consideration of the entropy of each word to capture informational relevance. We then introduce a regularized expectation-maximization (EM) algorithm model for parameters estimation, and integrate query commercial intent into the scoring function to boost overall click efficiency. Empirical evaluation shows that our model achieves better performance as compared to a well tuned classical language model and deliberated TFIDF-pLSI model (6% and 5% precision improvement at our operating point in production environment of 30% recall, and 5.3% and 6.3% AUC improvement), and performs superior to the KL Divergence language model for tail queries (0.5% nDCG improvement). Live traffic test shows over 2% CTR lift and 2.5% RPS lift as well.

Peer-to-Peer (P2P) applications have become increasingly popular in recent few years, which bring new challenges to network management and traffic engineering (TE). As basic input information, P2P traffic matrices are of significant importance for TE. Due to excessively high cost of direct measurement, a lot of studies aim at modeling and estimating general traffic matrices, but few focus on P2P traffic matrices. In this paper, we proposed a model to estimate P2P traffic matrices in networks. Important factors are considered, including the number of peers, the localization ratio of P2P traffic, and the distances among different networks. Here distance can be hop counts or geographic distance accordingly. To validate our model, we have evaluated the performance using both real P2P live steaming traces and file sharing application traces. Evaluation results show that the proposed model outperforms the other two typical models for general traffic matrices estimation, in terms of estimate errors. To the best of our knowledge, this is the first research on P2P traffic matrices estimation. P2P traffic matrices, derived from the model, can be applied to P2P traffic optimization and other TE fields.

The rise in peer-to-peer(P2P) networking has been tremendous in last several years. P2P traffic has significant impact on ISPs as it accounts for more than half of all traffic. Although many methods have been proposed to manage P2P traffic, little effort was spent on the deployment issues. In this paper, we have studied how to deploy P2P traffic cache devices in the backbone network to maximize the benefit of the ISP. A novel model is proposed to evaluate the benefits of deploying cache devices on different links. Guided by our model, two algorithms were developed to instruct the deployment of cache devices in the network. The experiment shows that deploying cached devices on less than 10% links can efficiently reduce the heavy load of the backbone network.