Online social networks (OSNs) serve today as a platform for information dissemination. At the same time, mobile devices provide ubiquitous network access through the cellular infrastructure. In this paper, we develop mechanisms for minimizing the peak load of the cellular network due to information cascades spreading on social media. First, we exploit the social ties for predicting information dissemination and we propose Proactive Seeding-a technique for minimizing the peak load of cellular networks. Much of such a load is due to information cascades spreading in social media, and we address it by proactively pushing (“seeding”) content to selected users before they actually request it. We develop a family of algorithms that take as input information primarily about: (i) cascades on the OSN, (ii) the background traffic load in the cellular network, and (iii) the local connectivity among mobiles; the algorithms then select which nodes to seed and when. We prove that Proactive Seeding is optimal when the prediction of information cascades is perfect. We perform simulations driven by traces from Twitter and cellular networks and we find that Proactive Seeding reduces the peak cellular load by 20-50 percent. Then, we exploit the fact that there is correlation between social ties and physical proximity and we combine Proactive Seeding with device-to-device communication to further reduce the peak load.

Understanding network structure and having access to realistic graphs plays a central role in computer and social networks research. In this paper, we propose a complete, practical methodology for generating graphs that resemble a real graph of interest. The metrics of the original topology we target to match are the joint degree distribution (JDD) and the degree-dependent average clustering coefficient (c̅(k)). We start by developing efficient estimators for these two metrics based on a node sample collected via either independence sampling or random walks. Then, we process the output of the estimators to ensure that the target metrics are realizable. Finally, we propose an efficient algorithm for generating topologies that have the exact target JDD and a c̅(k) close to the target. Extensive simulations using real-life graphs show that the graphs generated by our methodology are similar to the original graph with respect to, not only the two target metrics, but also a wide range of other topological metrics. Furthermore, our generator is order of magnitudes faster than state-of-the-art techniques.

Online social networks (OSNs) play an increasingly important role today in informing users about content. At the same time, mobile devices provide ubiquitous access to this content through the cellular infrastructure. In this paper, we exploit the fact that the interest in content spreads over OSNs, which makes it, to a certain extent, predictable. We propose Proactive Seeding-a technique for minimizing the peak load of cellular networks, by proactively pushing (“seeding”) content to selected users before they actually request it. We develop a family of algorithms that take as input information primarily about (i) cascades on the OSN and possibly about (ii) the background traffic load in the cellular network and (iii) the local connectivity among mobiles; the algorithms then select which nodes to seed and when. We prove that Proactive Seeding is optimal when the prediction of information cascades is perfect. In realistic simulations, driven by traces from Twitter and cellular networks, we find that Proactive Seeding reduces the peak cellular load by 20%-50%. Finally, we combine Proactive Seeding with techniques that exploit local mobile-to-mobile connections to further reduce the peak load.

Our objective is to sample the node set of a large unknown graph via crawling, to accurately estimate a given metric of interest. We design a random walk on an appropriately defined weighted graph that achieves high efficiency by preferentially crawling those nodes and edges that convey greater information regarding the target metric. Our approach begins by employing the theory of stratification to find optimal node weights, for a given estimation problem, under an independence sampler. While optimal under independence sampling, these weights may be impractical under graph crawling due to constraints arising from the structure of the graph. Therefore, the edge weights for our random walk should be chosen so as to lead to an equilibrium distribution that strikes a balance between approximating the optimal weights under an independence sampler and achieving fast convergence. We propose a heuristic approach (stratified weighted random walk, or S-WRW) that achieves this goal, while using only limited information about the graph structure and the node properties. We evaluate our technique in simulation, and experimentally, by collecting a sample of Facebook college users. We show that S-WRW requires 13-15 times fewer samples than the simple re-weighted random walk (RW) to achieve the same estimation accuracy for a range of metrics.

State-of-the-art techniques for probability sampling of users of online social networks (OSNs) are based on random walks on a single social relation (typically friendship). While powerful, these methods rely on the social graph being fully connected. Furthermore, the mixing time of the sampling process strongly depends on the characteristics of this graph. In this paper, we observe that there often exist other relations between OSN users, such as membership in the same group or participation in the same event. We propose to exploit the graphs these relations induce, by performing a random walk on their union multigraph. We design a computationally efficient way to perform multigraph sampling by randomly selecting the graph on which to walk at each iteration. We demonstrate the benefits of our approach through (i) simulation in synthetic graphs, and (ii) measurements of Last.fm- an Internet website for music with social networking features. More specifically, we show that multigraph sampling can obtain a representative sample and faster convergence, even when the individual graphs fail, i.e., are disconnected or highly clustered.

Our goal in this paper is to develop a practical framework for obtaining a uniform sample of users in an online social network (OSN) by crawling its social graph. Such a sample allows to estimate any user property and some topological properties as well. To this end, first, we consider and compare several candidate crawling techniques. Two approaches that can produce approximately uniform samples are the Metropolis-Hasting random walk (MHRW) and a re-weighted random walk (RWRW). Both have pros and cons, which we demonstrate through a comparison to each other as well as to the "ground truth." In contrast, using Breadth-First-Search (BFS) or an unadjusted Random Walk (RW) leads to substantially biased results. Second, and in addition to offline performance assessment, we introduce online formal convergence diagnostics to assess sample quality during the data collection process. We show how these diagnostics can be used to effectively determine when a random walk sample is of adequate size and quality. Third, as a case study, we apply the above methods to Facebook and we collect the first, to the best of our knowledge, representative sample of Facebook users. We make it publicly available and employ it to characterize several key properties of Facebook.

We consider a transmission of a delay-sensitive data stream (e.g., a video call) from a single source to a single destination. The reliability of this transmission may suffer from bursty packet losses - the predominant type of failures in today's Internet. An effective and well studied solution to this problem is to protect the data by a Forward Error Correction (FEC) code and send the FEC packets over multiple paths. In this paper, we show that the loss rate of such a classic multipath FEC scheme can often be significantly reduced, while keeping the total transmission rate and delay unchanged. Our key observation is that the propagation times on the available paths often significantly differ, usually by 10-100ms. We propose to exploit these differences by appropriate packet scheduling that we call `Spread'. We evaluate our solution with a precise, analytical formulation and trace-driven simulations. Our studies show that Spread substantially outperforms the state-of-the-art solutions. It typically achieves two- to five-fold improvement (reduction) in the effective loss rate. Or conversely, keeping the same level of effective loss rate, Spread significantly decreases the observed delays and helps fighting the delay jitter.

Breadth First Search (BFS) is a widely used approach for sampling large graphs. However, it has been empirically observed that BFS sampling is biased toward high-degree nodes, which may strongly affect the measurement results. In this paper, we quantify and correct the degree bias of BFS. First, we consider a random graph RG(pk) with an arbitrary degree distribution pk. For this model, we calculate the node degree distribution expected to be observed by BFS as a function of the fraction f of covered nodes. We also show that, for RG(pk), all commonly used graph traversal techniques (BFS, DFS, Forest Fire, Snowball Sampling, RDS) have exactly the same bias. Next, we propose a practical BFS-bias correction procedure that takes as input a collected BFS sample together with the fraction f. Our correction technique is exact (i.e., leads to unbiased estimation) for RG(pk). Furthermore, it performs well when applied to a broad range of Internet topologies and to two large BFS samples of Facebook and Orkut networks.

We present Netscope, a tomographic technique that infers the loss rates of network links from unicast end-to-end measurements. Netscope uses a novel combination of first- and second-order moments of end-to-end measurements to identify and characterize the links that cannot be (accurately) characterized through existing practical tomographic techniques. Using both analytical and experimental tools, we show that Netscope enables scalable, accurate link-loss inference: in a simulation scenario involving 4000 links, 20% of them lossy, Netscope correctly identifies 94% of the lossy links with a false positive rate of 16%-a significant improvement over the existing alternatives. Netscope is robust in the sense that it requires no parameter tuning, moreover its advantage over the alternatives widens when the number of lossy links increases. We also validate Netscope's performance on an "Internet tomographer" that we deployed on an overlay of 400 PlanetLab nodes.

With more than 250 million active users, Facebook (FB) is currently one of the most important online social networks. Our goal in this paper is to obtain a representative (unbiased) sample of Facebook users by crawling its social graph. In this quest, we consider and implement several candidate techniques. Two approaches that are found to perform well are the Metropolis-Hasting random walk (MHRW) and a re-weighted random walk (RWRW). Both have pros and cons, which we demonstrate through a comparison to each other as well as to the "ground-truth" (UNI - obtained through true uniform sampling of FB userIDs). In contrast, the traditional Breadth-First-Search (BFS) and Random Walk (RW) perform quite poorly, producing substantially biased results. In addition to offline performance assessment, we introduce online formal convergence diagnostics to assess sample quality during the data collection process. We show how these can be used to effectively determine when a random walk sample is of adequate size and quality for subsequent use (i.e., when it is safe to cease sampling). Using these methods, we collect the first, to the best of our knowledge, unbiased sample of Facebook. Finally, we use one of our representative datasets, collected through MHRW, to characterize several key properties of Facebook.

We consider a transmission of a delay-sensitive data stream from a single source to a single destination. The reliability of this transmission may suffer from bursty packet losses - the predominant type of failures in today's Internet. An effective and well studied solution to this problem is to protect the data by a forward error correction (FEC) code and send the FEC packets over multiple paths. In this paper we show that the performance of such a multipath FEC scheme can often be further improved. Our key observation is that the propagation times on the available paths often significantly differ, usually by 10-100ms. We propose to exploit these differences by appropriate packet scheduling that we call 'Spread'. We evaluate our solution with a precise, analytical formulation and trace-driven simulations. Our studies show that Spread substantially outperforms the state-of-the-art solutions. It typically achieves two- to five-fold improvement (reduction) in the effective loss rate. Or conversely, keeping the same level of effective loss rate, Spread significantly decreases the observed delays and helps fighting the delay jitter.

Failure restoration at the IP layer in IP-over-WDM networks requires to map the IP topology on the WDM topology in such a way that a failure at the WDM layer leaves the IP topology connected. Such a mapping is called survivable. Finding a survivable mapping is known to be NP-complete, making it impossible in practice to assess the existence or absence of such a mapping for large networks, (i) we first introduce a new concept of piecewise survivability, which makes the problem much easier in practice (although still NP-complete), and allows us to formally prove that a given survivable mapping does or does not exist, (ii) secondly, we show how to trace the vulnerable areas in the topology, and how to strengthen them to enable a survivable mapping, (iii) thirdly, we give an efficient and scalable algorithm that finds a survivable mapping. In contrast to the heuristics proposed in the literature to date, our algorithm exhibits a number of provable properties (e.g., it guarantees the piecewise survivability) that are crucial for (i) and (ii)

Failure restoration at the IP layer in IP-over-WDM networks requires to map the IP topology on the WDM topology in such a way that a failure at the WDM layer leaves the IP topology connected. Such a mapping is called survivable. Finding a survivable mapping is known to be NP-complete E. Modiano et al., (2002), making it impossible in practice to assess the existence or absence of such a mapping for large networks, (i) We first introduce a new concept of piecewise survivability, which makes the problem much easier, and allows us to formally prove that a given survivable mapping does or does not exist, (ii) Secondly, we show how to trace the vulnerable areas in the topology, and how to strengthen them to enable a survivable mapping, (iii) Thirdly, we give an efficient and scalable algorithm that finds a survivable mapping. In contrast to the heuristics proposed in the literature to date, our algorithm exhibits a number of provable properties that are crucial for (i) and (ii). We consider both link and node failures at the physical layer.

We develop a fast and efficient algorithm that finds a survivable (i.e., robust to single fiber failures) mapping of IP topology on the mesh of fibers in IP-over-WDM networks; we call it SMART. A number of algorithms solving this problem can be found in the literature. Since ILP solutions are highly complex, many heuristics were proposed. They usually start with some initial mapping and then try to gradually improve it. This involves the evaluation of the entire topology at each iteration, which is costly for large topologies. We propose a different approach. The SMART algorithm breaks down the task into a set of independent and very simple subtasks. The combination of solutions of these subtasks is a survivable mapping. This is why SMART is orders of magnitude faster than other proposals, especially when dealing with large topologies. We also extend the SMART algorithm to obtain a mapping resilient to fiber span failures, node failures and double-link failures. Finally, we show that the scalability of the standard heuristic approaches is additionally limited (contrary to SMART) when applied to double-link failures.