Daniel R. Figueiredo

Modeling peer-peer file sharing systems

Peer-peer networking has recently emerged as a new paradigm for building distributed networked applications. We develop simple mathematical models to explore and illustrate fundamental performance issues of peer-peer file sharing systems. The modeling framework introduced and the corresponding solution methods are flexible enough to accommodate different characteristics of such systems. Through the specification of model parameters, we apply our framework to three different peer-peer architectures: centralized indexing, distributed indexing with flooded queries, and distributed indexing with hashing directed queries. Using our model, we investigate the effects of system scaling, freeloaders, file popularity and availability on system performance. In particular, we observe that a system with distributed indexing and flooded queries cannot exploit the full capacity of peer-peer systems. We further show that peer-peer file sharing systems can tolerate a significant number of freeloaders without suffering much performance degradation. In many cases, freeloaders can benefit from the available spare capacity of peer-peer systems and increase overall system throughput. Our work shows that simple models coupled with efficient solution methods can be used to understand and answer questions related to the performance of peer-peer file sharing systems.

Characterizing and Detecting Skype-Relayed Traffic

Networked application developers have recently started to use end-users’ computers as relay nodes – application instances that also act as bridges between pairs of hosts running the same application. Relay nodes can bring costs to both users and network operators, at least in terms of increased bandwidth consumption. An interesting problem is to characterize the nature of relayed traffic and to detect its presence in the network. This paper focuses on characterizing and detecting relayed traffic generated by Skype, a popular voice over IP application that uses relays. Our technique relies solely on flow-level properties rather than on applicationor protocol-specific information. Using two different controlled experimental environments we generate and collect a large amount of Skype-relayed traffic. We propose several metrics to characterize the nature of relayed traffic. These metrics together with the results obtained from the experimental characterization of Skype-relayed traffic are used to detect Skyperelayed traffic traversing the access point of a large network. We show that the metrics proposed can be used to reliably detect Skype-relayed traffic. Finally, we believe the metrics proposed could be applied more broadly in the detection of relayed traffic generated by other multimedia applications.

On the hierarchical structure of the logical Internet graph

The study of the Internet topology has recently received much attention from the research community. In particular, the observation that the network graph has interesting properties, such as power laws, that might be explored in a myriad of ways. Most of the work in characterizing the Internet graph is based on the physical network graph, i.e., the connectivity graph. In this paper we investigate how logical relationships between nodes of the AS graph can be used to gain insight to its structure. We characterize the logical graph using various metrics and identify the presence of power laws in the number of customers that a provider has. Using these logical relationships we define a structural model of the AS graph. The model highlights the hierarchical nature of logical relationships and the preferential connection to larger providers. We also investigate the consistency of this model over time and observe interesting properties of the hierarchical structure.

Incentives to promote availability in peer-to-peer anonymity systems

Peer-to-peer (P2P) anonymous communication systems are vulnerable to free-riders, peers that use the system while providing little or no service to others and whose presence limits the strength of anonymity as well as the efficiency of the system. Free-riding can be addressed by building explicit incentive mechanisms into system protocols to promote two distinct aspects of cooperation among peers-compliance with the protocol specification and the availability of peers to serve others. In this paper we study the use of payments to implement an incentive mechanism that attaches a real monetary cost to low availability. Through a game theoretic analysis, we evaluate the effectiveness of such an incentive, finding that peer availability can be significantly increased through the introduction of payments under many conditions. We also demonstrate how a payment-based incentive that preserves anonymity can be implemented and integrated with a popular class of P2P anonymity systems.

struc2vec : Learning Node Representations from Structural Identity

Structural identity is a concept of symmetry in which network nodes are identified according to the network structure and their relationship to other nodes. Structural identity has been studied in theory and practice over the past decades, but only recently has it been addressed with representational learning techniques. This work presents struc2vec, a novel and flexible framework for learning latent representations for the structural identity of nodes. struc2vec uses a hierarchy to measure node similarity at different scales, and constructs a multilayer graph to encode structural similarities and generate structural context for nodes. Numerical experiments indicate that state-of-the-art techniques for learning node representations fail in capturing stronger notions of structural identity, while struc2vec exhibits much superior performance in this task, as it overcomes limitations of prior approaches. As a consequence, numerical experiments indicate that struc2vec improves performance on classification tasks that depend more on structural identity.

On the autocorrelation structure of TCP traffic

The statistical characteristics of network traffic--in particular the observation that it can exhibit long range dependence--have received considerable attention from the research community over the past few years. In addition, the recent claims that the TCP protocol can generate traffic with long range dependent behavior has also received much attention. Contrary to the latter claims, in this paper we show that the TCP protocol can generate traffic with correlation structures that spans only an analytically predictable finite range of time-scales. We identify and analyze separately the two mechanisms within TCP that are responsible for this scaling behavior: timeouts and congestion avoidance. We provide analytical models for both mechanisms that, under the proper loss probabilities, accurately predict the range in time-scales and the strength of the sustained correlation structure of the traffic sending rate of a single TCP source. We also analyze an existing comprehensive model of TCP that accounts for both mechanisms and show that TCP itself exhibits a predictable finite range of time-scales under which traffic presents sustained correlations. Our claims and results are derived from Markovian models that are supported by simulations. We note that traffic generated by TCP can be misinterpreted to have long range dependence, but that long range dependence is not possible due to inherent finite time-scales of the mechanisms of TCP.

Efficient mechanisms for recovering voice packets in the Internet

Multimedia applications, such as voice transmission, have increasingly been used over the Internet. However, there are still many issues under investigation, related to the quality of audio delivered, such as the reduction of the jitter and the loss of audio packets. Concerning audio packet loss, many recovery mechanisms have been proposed to improve the audio quality. However, the efficiency of these mechanisms is largely dependent on the loss process in the network. In this paper we study the packet loss process of audio streams aiming at comparing different recovery mechanisms. We also propose a new packet recovery mechanism and show that its efficiency can be significantly better than others proposed in the literature.

A Bayesian method for matching two similar graphs without seeds

Approximate graph matching (AGM) refers to the problem of mapping the vertices of two structurally similar graphs, which has applications in social networks, computer vision, chemistry, and biology. Given its computational cost, AGM has mostly been limited to either small graphs (e.g., tens or hundreds of nodes), or to large graphs in combination with side information beyond the graph structure (e.g., a seed set of pre-mapped node pairs). In this paper, we cast AGM in a Bayesian framework based on a clean definition of the probability of correctly mapping two nodes, which leads to a polynomial time algorithm that does not require side information. Node features such as degree and distances to other nodes are used as fingerprints. The algorithm proceeds in rounds, such that the most likely pairs are mapped first; these pairs subsequently generate additional features in the fingerprints of other nodes. We evaluate our method over real social networks and show that it achieves a very low matching error provided the two graphs are sufficiently similar. We also evaluate our method on random graph models to characterize its behavior under various levels of node clustering.

On TCP and self-similar traffic

We re-examine the same TCP trace that was used by Veres et al. [A. Veres, M. Boda, The chaotic nature of TCP congestion control, in: Proceedings of the IEEE INFOCOM, 2000] to claim that TCP creates self-similar traffic. A careful reassessment of their data analysis shows that this claim is not justified and suggests that the TCP trace in question is not consistent with (asymptotic second-order) self-similarity or long-range dependence (LRD). We illustrate the reasons that led to the claim in [A. Veres, M. Boda, The chaotic nature of TCP congestion control, in: Proceedings of the IEEE INFOCOM, 2000] and provide some practical guidelines for assessing a statistical characteristic of trace data such as LRD that is defined in strictly asymptotic terms. Our conclusion is in full agreement with the findings obtained from analyzing a much longer TCP trace (resulting from repeating the same simulation as in [A. Veres, M. Boda, The chaotic nature of TCP congestion control, in: Proceedings of the IEEE INFOCOM, 2000], but running it for a longer period) and with analytical results derived from a detailed Markovian model of TCP. These results show that the traffic generated by a long-lived TCP connection, while exhibiting pronounced correlations over a predictable finite range of time-scales, cannot be (asymptotically second-order) self-similar or exhibit LRD. Our work serves as a reminder of the importance of careful trace analysis and detailed examination (and cross-validation) of alternative explanations when establishing or characterizing the generality of any particular finding about Internet traffic.

An evolutionary game-theoretic approach to congestion control

This paper investigates a system where a set of users sharing a bottleneck link must choose the transmission rate at which multimedia traffic is received. Users are assumed to be self-regarding and make their decisions with the sole goal of maximizing their perceived quality. We are interested in the dynamic process by which users adapt their data rates and the convergence of this process to equilibria. We propose a novel two-layer model to represent this system: the upper layer is an evolutionary game-theoretic model that captures how users adapt their rates; the lower layer model captures the network performance and the quality perceived by the users. Using the model proposed, we demonstrate analytically and numerically several interesting properties of the system equilibria. In particular, we establish the relationship between system states that have non-negligible steady state probabilities and Nash equilibria of the induced game.