Tuan Anh Trinh

Scafida: a scale-free network inspired data center architecture

Data centers have a crucial role in current Internet architecture supporting content-centric networking. State-of-the-art data centers have different architectures like fat-tree, DCell, or BCube. However, their architectures share a common property: symmetry. Due to their symmetric nature, a tricky point with these architectures is that they are hard to be extended in small quantities. Contrary to state-of-the-art data center architectures, we propose an asymmetric data center topology generation method called Scafida inspired by scale-free networks; these data centers have not only small diameters and high fault tolerance, inherited by scale-free networks, but can also be scaled in smaller and less homogenous increments. We extend the original scale-free network generation algorithm of Barabasi and Albert to meet the physical constraints of switches and routers. Despite the fact that our method artificially limits the node degrees in the network, our data center architectures keep the preferable properties of scale-free networks. Based on extensive simulations we present preliminary results that are promising regarding the error tolerance, scalability, and flexibility of the architecture.

How can architecture help to reduce energy consumption in data center networking

Reducing the power consumption of data centers has recently been received much attention from the research community and the industry alike. It is because the benefits are manifold: for the data centers themselves and towards the environment. A number of issues concerning with the consumption of the servers or the switches have been investigated. However, the energy requirement of the data center architecture has not yet been sufficiently addressed. In this paper, we address this issue and make two key contributions. Before delving into the details of possible solutions for reducing power of existing data center architectures, we first characterize the impact of architectural parameters on the power consumption of data centers by presenting preliminary results on the power consumption of state-of-the-art data center structures, namely BCube [8], DCell [9], and Fat-tree [1, 11, 7]. Secondly, based on the insights and lessons learnt from the analysis, we present our vision on possible practical solutions to reduce the energy usage of data center networking.

A Comprehensive Performance Analysis of Random Early Detection Mechanism

One of the most promising active queue management schemes being proposed for deployment in the Internet is the Random Early Detection (RED) scheme. However, research results on RED performance are highly mixed, especially in the field of tuning its parameters. In this paper, a comprehensive performance analysis of RED is presented. We revisit some features in RED and study them in greater detail. We point out that RED, in general, does not possess proportional loss between flows as claimed and widely adopted in previous research. We suggest the generalization of the PASTA property and give a proof for TCP flows. We also evaluate the performance of the Exponential Weighted Moving Average (EWMA) algorithm in RED. We find that EWMA in RED is an unbiased estimator of the average queue-length, regardless of the weighting value wq. We also point out the theoretical and practical limits of EWMA in RED. Finally, we propose the use of fuzzy EWMA to RED (fuzzy RED) to alleviate the inflexibility of RED tuning. We use simulations to evaluate the performance of fuzzy RED and compare it with other versions of RED. Our simulations show that, in the case of a high workload and a high level of variation, fuzzy RED, by tracking system variation in an on-line manner, improves RED performance in a number of important router-based metrics like packet loss rate, average queueing delay, link utilization, and global power.

Pricing internet access for disloyal users: a game-theoretic analysis

In this paper we investigate the impact of customer loyalty on the price competition between local Internet Service Providers who sell Internet access to end-users. The main contribution of this paper is threefold. First, we develop a repeated game, and show how cooperation between ISPs resulting in higher profits can be enforced through a threat strategy in the presence of customer loyalty. Second, we investigate the case of a differentiated customer population by introducing dual reservation values, and show how it leads to new, pure strategy Nash equilibra for a wide range of demand functions. Third, we develop two novel models for customer loyalty, along with a simulation tool that is capable of demonstrating the impact of the novel models. We argue that our findings can bring us closer to the understanding of economic interactions among ISPs and, at the same time, can motivate researchers to incorporate a finer-grained user behavior model involving customer loyalty in their investigations of such interactions.

A Game-Theoretic Analysis of TCP Vegas

We use the tools from game theory to understand the impacts of the inherent congestion pricing schemes in TCP Vegas as well as the problems of parameter setting of TCP Vegas on its performance. It is shown how these inherent pricing schemes result in a rate control equilibrium state that is a Nash equilibrium which is also a global optimum of the all-Vegas networks. On the other hand, if the TCP Vegas’ users are assumed to be selfish in terms of setting their desired number of backlogged packets in the buffers along their paths, then the network as a whole, in certain circumstances, would operate very inefficiently. This poses a serious threat to the possible deployment of Vegas-based TCP (such as FAST TCP) in the future Internet.

Free-scaling your data center

Abstract The increasing popularity of both small and large private clouds and expanding public clouds poses new requirements to data center (DC) architectures. First, DC architectures should be incrementally scalable allowing the creation of DCs of arbitrary size with consistent performance characteristics. Second, initial DC deployments should be incrementally expandable supporting small-scale upgrades without decreasing operation efficiency. A DC architecture possessing both properties satisfies the requirement of free-scaling . Recent work in DC design focuses on traditional performance and scalability characteristics, therefore resulting in symmetric topologies whose upgradability is coarse-grained at best. In our earlier work we proposed Scafida, an asymmetric, scale-free network inspired DC topology which scales incrementally and has favorable structural characteristics. In this paper, we build on Scafida and propose a full-fledged DC architecture achieving free-scaling called FScafida . Our main contribution is threefold. First, we propose an organic expansion algorithm for FScafida; this combined with Scafida’s flexible original design results in a freely scalable architecture. Second, we introduce the Effective Source Routing mechanism that provides near-shortest paths, multi-path and multicast capability, and low signaling overhead by exploiting the benefits of the FScafida topology. Third, we show based on extensive simulations and a prototype implementation that FScafida is capable of handling the traffic patterns characteristic of both enterprise and cloud data centers, tolerates network equipment failures to a high degree, and allows for high bisection bandwidth.

Energy Efficiency of Data Centers

As the power consumption has a significant and continuously increasing part of the operational expenses of data centers, energy efficient data center networking has received special attention from the academic and industrial research community recently. The complex design of data centers provides several directions toward more energy efficiency, including the consumption of servers and network equipments. We address the issue of power consumption of data centers from a higher-level point of view by analyzing the energy efficiency of data center architectures. We review the state-of-the-art data center architectures including BCube, DCell, fat-tree, and Scafida and present evaluate their energy-efficiency quantitatively. In addition, the trade-off between the power consumption and the performance of the data center is investigated. Next, other aspects of data centers’ energy efficiency is revealed including thermal control techniques, energy management systems. Finally, a standard data center energy efficiency metric called PUE (Power Usage Effectiveness) is presented.

A Research Framework for Analyzing High Speed Transport Protocols Based on Control-theory

In the recent years, several transport protocols have been proposed for very high bandwidth-delay product networks. However, little is known about the performance of these new mechanisms as well as the interaction with other elements of the network (such as the RED queue management). On the other hand, the control-theoretic approach has proved to be a very useful tool in order to get analytical insight into the performance of congestion control algorithms. In this paper, a control-theoretic research framework is designed and implemented for analyzing high speed transport protocol proposals in network environments with RED active queue management. As a case study, a comprehensive control-theoretic analysis of a promising proposal, namely the HighSpeed TCP is provided. The main contributions of this paper are the following. First, we provide a fluid-flow model for HighSpeed TCP/RED networks. Second, a comprehensive and systematic implementation methodology is described in detail, and a Simulink-based framework is designed for analyzing fluid-based models. Third, we derive a stability condition for HighSpeed TCP/RED networks. The flow-level results are validated by packet-level simulations conducted in Ns-2. Finally, an extension of the framework is presented that makes it capable of describing the interaction of different transport protocols.

A User-Influenced Pricing Mechanism for Internet Access

Proper pricing schemes are vital components to the continuing success of the Internet. In this paper, we propose a new pricing scheme for Internet access called user-influenced pricing. Our main contribution is threefold: first, we show how user-influenced pricing can provide the ISP with calculable revenues, while giving the users a chance to lower their costs via voting for their preferred pricing scheme. Second, we develop a cooperative weighted voting game which models the decision-making process, and we derive equilibrium solutions to analyze possible outcomes of the vote. Third, we investigate the distribution of power and we show that users with medium generated traffic volume are pivotal to the outcome. Finally, we discuss the practical feasibility of the proposed mechanism regarding user population, revenue planning and charging.

Path‐based Multipath Protection: resilience using multiple paths

The Multipath Protection (MPP) network resilience methods route and protect the upcoming demand over multiple paths. Although these schemes are able to allocate resources efficiently in the network, the length of the paths that the packets traverse cannot be controlled. Accordingly, to control the latency of the connections, we reuse the idea of MPP when we propose the Path-based Multipath Protection (PMPP) scheme. By controlling the maximal length of the paths, we increase the availability as well, because on shorter paths, fewer failures happen. The contribution of our work is threefold. First, we give the linear programming formulation of the PMPP method that does not have any integer variables resulting from polynomial runtime, similar to the former MPP methods. Second, we present extensive simulation results including the blocking ratios, the properties of the paths and the availabilities. Finally, we compare the PMPP resilience scheme with the original and improved versions of MPP method on the basis of several metrics. Copyright