Zongpeng Li

Youtube traffic characterization: a view from the edge

This paper presents a traffic characterization study of the popular video sharing service, YouTube. Over a three month period we observed almost 25 million transactions between users on an edge network and YouTube, including more than 600,000 video downloads. We also monitored the globally popular videos over this period of time. In the paper we examine usage patterns, file properties, popularity and referencing characteristics, and transfer behaviors of YouTube, and compare them to traditional Web and media streaming workload characteristics. We conclude the paper with a discussion of the implications of the observed characteristics. For example, we find that as with the traditional Web, caching could improve the end user experience, reduce network bandwidth consumption, and reduce the load on YouTubes core server infrastructure. Unlike traditional Web caching, Web 2.0 provides additional meta-data that should be exploited to improve the effectiveness of strategies like caching.

A Cross-Layer Optimization Framework for Multihop Multicast in Wireless Mesh Networks

The optimal and distributed provisioning of high throughput in mesh networks is known as a fundamental but hard problem. The situation is exacerbated in a wireless setting due to the interference among local wireless transmissions. In this paper, we propose a cross-layer optimization framework for throughput maximization in wireless mesh networks, in which the data routing problem and the wireless medium contention problem are jointly optimized for multihop multicast. We show that the throughput maximization problem can be decomposed into two subproblems: a data routing subproblem at the network layer, and a power control subproblem at the physical layer with a set of Lagrangian dual variables coordinating interlayer coupling. Various effective solutions are discussed for each subproblem. We emphasize the network coding technique for multicast routing and a game theoretic method for interference management, for which efficient and distributed solutions are derived and illustrated. Finally, we show that the proposed framework can be extended to take into account physical-layer wireless multicast in mesh networks

On achieving optimal throughput with network coding

With the constraints of network topologies and link capacities, achieving the optimal end-to-end throughput in data networks has been known as a fundamental but computationally hard problem. In this paper, we seek efficient solutions to the problem of achieving optimal throughput in data networks, with single or multiple unicast, multicast and broadcast sessions. Although previous approaches lead to solving NP-complete problems, we show the surprising result that, facilitated by the recent advances of network coding, computing the strategies to achieve the optimal end-to-end throughput can be performed in polynomial time. This result holds for one or more communication sessions, as well as in the overlay network model. Supported by empirical studies, we present the surprising observation that in most topologies, applying network coding may not improve the achievable optimal throughput; rather, it facilitates the design of significantly more efficient algorithms to achieve such optimality.

On achieving maximum multicast throughput in undirected networks

The transmission of information within a data network is constrained by the network topology and link capacities. In this paper, we study the fundamental upper bound of information dissemination rates with these constraints in undirected networks, given the unique replicable and encodable properties of information flows. Based on recent advances in network coding and classical modeling techniques in flow networks, we provide a natural linear programming formulation of the maximum multicast rate problem. By applying Lagrangian relaxation on the primal and the dual linear programs (LPs), respectively, we derive a) a necessary and sufficient condition characterizing multicast rate feasibility, and b) an efficient and distributed subgradient algorithm for computing the maximum multicast rate. We also extend our discussions to multiple communication sessions, as well as to overlay and ad hoc network models. Both our theoretical and simulation results conclude that, network coding may not be instrumental to achieve better maximum multicast rates in most cases; rather, it facilitates the design of significantly more efficient algorithms to achieve such optimality.

The flattening internet topology: natural evolution, unsightly barnacles or contrived collapse?

In this paper we collect and analyze traceroute measurements to show that large content providers (e.g., Google, Microsoft, Yahoo!) are deploying their own wide-area networks, bringing their networks closer to users, and bypassing Tier-1 ISPs on many paths. This trend, should it continue and be adopted by more content providers, could flatten the Internet topology, and may result in numerous other consequences to users, Internet Service Providers (ISPs), content providers, and network researchers.

Dynamic resource provisioning in cloud computing: A randomized auction approach

This work studies resource allocation in a cloud market through the auction of Virtual Machine (VM) instances. It generalizes the existing literature by introducing combinatorial auctions of heterogeneous VMs, and models dynamic VM provisioning. Social welfare maximization under dynamic resource provisioning is proven NP-hard, and modeled with a linear integer program. An efficient α-approximation algorithm is designed, with α ~ 2.72 in typical scenarios. We then employ this algorithm as a building block for designing a randomized combinatorial auction that is computationally efficient, truthful in expectation, and guarantees the same social welfare approximation factor α. A key technique in the design is to utilize a pair of tailored primal and dual LPs for exploiting the underlying packing structure of the social welfare maximization problem, to decompose its fractional solution into a convex combination of integral solutions. Empirical studies driven by Google Cluster traces verify the efficacy of the randomized auction.

Strategyproof auctions for balancing social welfare and fairness in secondary spectrum markets

Secondary spectrum access is emerging as a promising approach for mitigating the spectrum scarcity in wireless networks. Coordinated spectrum access for secondary users can be achieved using periodic spectrum auctions. Recent studies on such auction design mostly neglect the repeating nature of such auctions, and focus on greedily maximizing social welfare. Such auctions can cause subsets of users to experience starvation in the long run, reducing their incentive to continue participating in the auction. It is desirable to increase the diversity of users allocated spectrum in each auction round, so that a trade-off between social welfare and fairness is maintained. We study truthful mechanisms towards this objective, for both local and global fairness criteria. For local fairness, we introduce randomization into the auction design, such that each user is guaranteed a minimum probability of being assigned spectrum. Computing an optimal, interference-free spectrum allocation is NP-Hard; we present an approximate solution, and tailor a payment scheme to guarantee truthful bidding is a dominant strategy for all secondary users. For global fairness, we adopt the classic maxmin fairness criterion. We tailor another auction by applying linear programming techniques for striking the balance between social welfare and max-min fairness, and for finding feasible channel allocations. In particular, a pair of primal and dual linear programs are utilized to guide the probabilistic selection of feasible allocations towards a desired tradeoff in expectation.

A Constant Bound on Throughput Improvement of Multicast Network Coding in Undirected Networks

Recent research in network coding shows that, joint consideration of both coding and routing strategies may lead to higher information transmission rates than routing only. A fundamental question in the field of network coding is: how large can the throughput improvement due to network coding be? In this paper, we prove that in undirected networks, the ratio of achievable multicast throughput with network coding to that without network coding is bounded by a constant ratio of 2, i.e., network coding can at most double the throughput. This result holds for any undirected network topology, any link capacity configuration, any multicast group size, and any source information rate. This constant bound 2 represents the tightest bound that has been proved so far in general undirected settings, and is to be contrasted with the unbounded potential of network coding in improving multicast throughput in directed networks.

Moving Big Data to The Cloud: An Online Cost-Minimizing Approach

Cloud computing, rapidly emerging as a new computation paradigm, provides agile and scalable resource access in a utility-like fashion, especially for the processing of big data. An important open issue here is to efficiently move the data, from different geographical locations over time, into a cloud for effective processing. The de facto approach of hard drive shipping is not flexible or secure. This work studies timely, cost-minimizing upload of massive, dynamically-generated, geo-dispersed data into the cloud, for processing using a MapReduce-like framework. Targeting at a cloud encompassing disparate data centers, we model a cost-minimizing data migration problem, and propose two online algorithms: an online lazy migration (OLM) algorithm and a randomized fixed horizon control (RFHC) algorithm , for optimizing at any given time the choice of the data center for data aggregation and processing, as well as the routes for transmitting data there. Careful comparisons among these online and offline algorithms in realistic settings are conducted through extensive experiments, which demonstrate close-to-offline-optimum performance of the online algorithms.

Scaling social media applications into geo-distributed clouds

Federation of geo-distributed cloud services is a trend in cloud computing that, by spanning multiple data centers at different geographical locations, can provide a cloud platform with much larger capacities. Such a geo-distributed cloud is ideal for supporting large-scale social media applications with dynamic contents and demands. Although promising, its realization presents challenges on how to efficiently store and migrate contents among different cloud sites and how to distribute user requests to the appropriate sites for timely responses at modest costs. These challenges escalate when we consider the persistently increasing contents and volatile user behaviors in a social media application. By exploiting social influences among users, this paper proposes efficient proactive algorithms for dynamic, optimal scaling of a social media application in a geo-distributed cloud. Our key contribution is an online content migration and request distribution algorithm with the following features: 1) future demand prediction by novelly characterizing social influences among the users in a simple but effective epidemic model; 2) one-shot optimal content migration and request distribution based on efficient optimization algorithms to address the predicted demand; and 3) a Δ(t)-step look-ahead mechanism to adjust the one-shot optimization results toward the offline optimum. We verify the effectiveness of our online algorithm by solid theoretical analysis, as well as thorough comparisons to ready algorithms including the ideal offline optimum, using large-scale experiments with dynamic realistic settings on Amazon Elastic Compute Cloud (EC2).