ArchiveSpace Name Record
Publication Search Results
Now showing 1 - 10 of 10
ItemML@GT Lab presents LAB LIGHTNING TALKS 2020( 2020-12-04) AlRegib, Ghassan ; Chau, Duen Horng ; Chava, Sudheer ; Cohen, Morris B. ; Davenport, Mark A. ; Desai, Deven ; Dovrolis, Constantine ; Essa, Irfan ; Gupta, Swati ; Huo, Xiaoming ; Kira, Zsolt ; Li, Jing ; Maguluri, Siva Theja ; Pananjady, Ashwin ; Prakash, B. Aditya ; Riedl, Mark O. ; Romberg, Justin ; Xie, Yao ; Zhang, XiuweiLabs affiliated with the Machine Learning Center at Georgia Tech (ML@GT) will have the opportunity to share their research interests, work, and unique aspects of their lab in three minutes or less to interested graduate students, Georgia Tech faculty, and members of the public. Participating labs include: Yao’s Group - Yao Xie, H. Milton Stewart School of Industrial Systems and Engineering (ISyE); Huo Lab - Xiaoming Huo, ISyE; LF Radio Lab – Morris Cohen, School of Electrical Computing and Engineering (ECE); Polo Club of Data Science – Polo Chau, CSE; Network Science – Constantine Dovrolis, School of Computer Science; CLAWS – Srijan Kumar, CSE; Control, Optimization, Algorithms, and Randomness (COAR) Lab – Siva Theja Maguluri, ISyE; Entertainment Intelligence Lab and Human Centered AI Lab – Mark Riedl, IC; Social and Language Technologies (SALT) Lab – Diyi Yang, IC; FATHOM Research Group – Swati Gupta, ISyE; Zhang's CompBio Lab – Xiuwei Zhang, CSE; Statistical Machine Learning - Ashwin Pananjady, ISyE and ECE; AdityaLab - B. Aditya Prakash, CSE; OLIVES - Ghassan AlRegib, ECE; Robotics Perception and Learning (RIPL) – Zsolt Kira, IC; Eye-Team - Irfan Essa, IC; and Mark Davenport, ECE.
ItemA Model of Interdomain Network Formation, Economics and Routing(Georgia Institute of Technology, 2009) Dhamdhere, Amogh ; Dovrolis, ConstantineThe Internet at the interdomain level is highly dynamic, as autonomous networks change their connectivity to optimize either monetary cost, profit and/or performance. Internet Service Providers (ISPs), for example, are mainly concerned with maximizing their profits, and they attempt to do so by changing their set of providers or peers. It is not well understood, however, what the properties of the resulting internetwork are, in terms of topology, economics and performance. In this paper, we propose ITER, a first-principles model of interdomain network formation that incorporates the effects of economics, interdomain traffic flow, geography, pricing/cost structures and interdomain routing policies. We use an agent-based computational method (treating networks as selfish agents) to find the equilibrium that results as each network uses a certain provider and peer selection strategy (such as “peer by traffic ratios” or “peer by necessity”). We study the properties of this equilibrium in terms of topology, traffic flow and economics. We also investigate the effect of factors such as the interdomain traffic matrix, geography, and customer preferences on the properties of the equilibrium network.
ItemOn the Congestion Responsiveness of Aggregate Internet Traffic: Open-loop vs Closed-loop Session Arrivals(Georgia Institute of Technology, 2006) Prasad, Ravi S. ; Dovrolis, ConstantineA traffic aggregate is congestion responsive if it reacts to network congestion by reducing its rate. The congestion responsiveness of Internet traffic has been largely attributed to TCP's congestion control. In this paper, we argue that congestion control for individual transfers is not sufficient to produce responsive aggregate traffic. The offered load at a network link is generated from users/applications that generate finite-length flows or groups of flows (sessions). We examine two session generation models. First, a closed-loop model where each user from a certain population can generate a new session only after the completion of her previous session. Second, an open-loop model where sessions arrive independently of previous sessions. These two models produce traffic with very different congestion responsiveness, even if each flow is controlled by TCP. We introduce two metrics to quantify the congestion responsiveness of a traffic aggregate, the throughput responsiveness and the flow rate responsiveness, and show that the closed-loop model results in congestion responsive traffic, while the open-loop model can lead to persistent overload and congestion collapse. We then measure the congestion responsiveness of the traffic at a university access link. These experiments show that both responsiveness metrics are close to zero, which explains why that link is often under persistent overload. We also present an estimation methodology to classify the traffic at a link as open-loop or closed-loop. Our measurements at a dozen of access and core links show that more than 70% of the traffic we analyzed follows the closed-loop model. This implies that a major reason for the congestion responsiveness of Internet traffic may be that most traffic reacts to congestion at the session generation layer.
ItemOn the Predictability of Large Transfer TCP Throughput(Georgia Institute of Technology, 2005) He, Qi ; Dovrolis, Constantine ; Ammar, Mostafa H.With the advent of overlay and peer-to-peer networks, Grid computing, and CDNs, network performance prediction becomes an essential task. Predicting the throughput of large TCP transfers, in particular, has attracted much attention. In this work, we focus on the design, empirical evaluation, and analysis of TCP throughput predictors for a broad class of applications. We first classify TCP throughput prediction techniques into two categories: Formula-Based (FB) and History-Based (HB). Within each class, we develop representative prediction algorithms, which we then evaluate empirically over the RON testbed. FB prediction relies on mathematical models that express the TCP throughput as a function of the characteristics of the network path (e.g., RTT, loss rate, available bandwidth). FB prediction does not rely on previous TCP transfers in the given path, and it can be performed with non-intrusive network measurements. We show, however, that the FB method is accurate only if the TCP transfer is window-limited to the point that it does not saturate the underlying path, and explain the main causes of the prediction errors. HB techniques predict the throughput of TCP flows from a time series of previous TCP throughput measurements on the same path, when such a history is available. We show that even simple HB predictors, such as Moving Average and Holt-Winters, using a history of limited and sporadic samples, can be quite accurate. On the negative side, HB predictors are highly path-dependent. Using simple queueing models, we explain the cause of such path dependencies based on two key factors: the load on the path, and the degree of statistical multiplexing.
ItemOn the Predictability of Large Transfer TcP Throughput(Georgia Institute of Technology, 2005) Dovrolis, Constantine ; Ammar, Mostafa H. ; He, QiWith the advent of overlay and peer-to-peer networks, Grid computing, and CDNs, network performance prediction becomes an essential task. Predicting the throughput of large TCP transfers, in particular, has attracted much attention. In this work, we focus on the design, empirical evaluation, and analysis of TCP throughput predictors for a broad class of applications. We first classify TCP throughput prediction techniques into two categories: Formula-Based (FB) and History-Based (HB). Within each class, we develop representative prediction algorithms, which we then evaluate empirically over the RON testbed. FB prediction relies on mathematical models that express the TCP throughput as a function of the characteristics of the network path (e.g., RTT, loss rate, available bandwidth). FB prediction does not rely on previous TCP transfers in the given path, and it can be performed with non-intrusive network measurements. We show, however, that the FB method is accurate only if the TCP transfer is window-limited to the point that it does not saturate the underlying path, and explain the main causes of the prediction errors. HB techniques predict the throughput of TCP flows from a time series of previous TCP throughput measurements on the same path, when such a history is available. We show that even simple HB predictors, such as Moving Average and Holt-Winters, using a history of limited and sporadic samples, can be quite accurate. On the negative side, HB predictors are highly pathdependent. Using simple queueing models, we explain the cause of such path dependencies based on two key factors: the load on the path, and the degree of statistical multiplexing.
ItemEnd-to-end Estimation of the Available Bandwidth Variation Range(Georgia Institute of Technology, 2005) Jain, Manish ; Dovrolis, ConstantineThe available bandwidth (avail-bw) of a network path is an important performance metric and its end-to-end estimation has recently received significant attention. Previous work focused on the estimation of the average avail-bw, ignoring the significant variability of this metric in different time scales. In this paper, we show how to estimate a given percentile of the avail-bw distribution at a user-specified time scale. If two estimated percentiles cover the bulk of the distribution (say 10% to 90%), the user can obtain a practical estimate for the avail-bw variation range. We present two estimation techniques. The first is iterative and non-parametric, meaning that it is more appropriate for very short time scales (typically less than 100ms), or in bottlenecks with limited flow multiplexing (where the avail-bw distribution may be non-Gaussian). The second technique is parametric, because it assumes that the avail-bw follows the Gaussian distribution, and it can produce an estimate faster because it is not iterative. The two techniques have been implemented in a measurement tool called Pathvar. Pathvar can track the avail-bw variation range within 10-20%, even under non-stationary conditions. We identify four factors that play a crucial role in the variation range of the avail-bw: traffic load, number of competing flows, rate of competing flows, and of course the measurement time scale. Finally, we present a new way to detect whether a probing rate is larger than the avail-bw, without relying on the fluid traffic assumption or on static thresholds.
ItemInterdomain Ingress Traffic Engineering through Optimized AS-Path Prepending(Georgia Institute of Technology, 2005) Dovrolis, Constantine ; Zegura, Ellen W. ; Gao, RuomeiIn Interdomain Ingress Traffic Engineering (INITE), a "target" Autonomous System (AS) aims to control the ingress link through which the traffic of one or more upstream source networks flows to the target network or to its customers. Currently, there are few methodologies for systematic INITE. In practice, ISPs often attempt to manipulate, mostly in a trial-and-error manner, the AS-Path length attribute of upstream routes through a simple technique known as prepending (or padding). In this paper, we focus on prepending and propose a polynomial-time algorithm (referred to as OPV) that determines the optimal padding for an upstream route at each ingress link of the target network. Specifically given a set of "elephant" source networks for a particular customer of the target network, and a set of maximum load constraints on the ingress links of the latter, OPV determines the minimum padding at each ingress link so that the load constraints are met, when it is feasible to do so. OPV requires as input an AS-Path length estimate from each source to each ingress link. We describe how to estimate this matrix, leveraging the BGP Looking Glass Servers that are abundant today for monitoring interdomain routing. To deal with unavoidable inaccuracies in the AS-Path length estimates, and also to compensate for the generally unknown BGP tie-breaking process in upstream networks, we develop a robust variation (RPV) of the OPV algorithm. We show that RPV manages to identify a padding vector that meets the given maximum load constraints, when it is feasible to do so, even in the presence of inaccurate AS-Path lengths and unknown BGP tie-breaking behavior.
ItemThe origin of TCP traffic burstiness in short time scales(Georgia Institute of Technology, 2004-02-20) Jiang, Hao ; Dovrolis, ConstantineInternet traffic exhibits multifaceted burstiness and correlation structure over a wide span of time scales. Previous work analyzed this structure in terms of heavy-tailed session characteristics, as well as TCP timeouts and congestion avoidance, in relatively long time scales. We focus on shorter scales, typically less than 100-1000 milliseconds. Our objective is to identify the actual mechanisms that are mostly responsible for creating bursty traffic in those scales. We show that TCP self-clocking, joint with queueing in the network, can shape the packet interarrivals of a TCP connection in a two-level ON-OFF pattern. This structure creates strong correlations and burstiness in time scales that extend up to the Round-Trip Time (RTT) of the connection, especially for bulk transfers that have a large bandwidth-delay product relative to their window size. Also, the aggregation of many flows, without rescaling their packet interarrivals, does not converge to a Poisson stream, as one might expect from classical superposition results. Instead, the burstiness in those scales can be significantly reduced by TCP pacing, depending however on the minimum pacing timer. Finally, we show that sub-RTT burstiness is important in queueing performance not only in moderate load conditions, as previously shown, but also in high loads when the bottleneck buffer size is relatively small.
ItemSocket Buffer Auto-Sizing for High-Performance Data Transfers(Georgia Institute of Technology, 2004) Prasad, Ravi S. ; Jain, Manish ; Dovrolis, ConstantineIt is often claimed that TCP is not a suitable transport protocol for data intensive Grid applications in high-performance networks. We argue that this is not necessarily the case. Without changing the TCP protocol, congestion control, or implementation, we show that an appropriately tuned TCP bulk transfer can saturate the available bandwidth of a network path. The proposed technique, called SOBAS, is based on automatic socket buffer sizing at the application layer. In non-congested paths, SOBAS limits the socket buffer size based on direct measurements of the received throughput and of the corresponding round-trip time. The key idea is that the send window should be limited, after the transfer has saturated the available bandwidth in the path, so that the transfer does not cause buffer overflows ("self-induced losses"). A difference with other socket buffer sizing schemes is that SOBAS does not require prior knowledge of the path characteristics, and it can be performed while the transfer is in progress. Experimental results in several high bandwidth-delay product paths show that SOBAS provides consistently a significant throughput increase (20% to 80%) compared to TCP transfers that use the maximum possible socket buffer size. We expect that SOBAS will be mostly useful for applications such as GridFTP in non-congested wide-area networks.
ItemThe TCP Bandwidth-Delay Product revisited: network buffering, cross traffic, and socket buffer auto-sizing(Georgia Institute of Technology, 2003) Jain, Manish ; Prasad, Ravi S. ; Dovrolis, ConstantineTCP is often blamed that it cannot use efficiently network paths with high Bandwidth-Delay Product (BDP). The BDP is of fundamental importance because it determines the required socket buffer size for maximum throughput. In this paper, we re-examine the BDP concept, considering the effects of network buffering and cross traffic on the `bandwidth' and `delay' characteristics of a path. We show that, with careful socket buffer sizing, a bulk TCP transfer can saturate a network path independent of the BDP or the available network buffers. In a non-congested path, there is a certain socket buffer size (which depends on the cross traffic type) that maximizes the throughput of a bulk TCP transfer. In a congested path, the TCP throughput is maximized when the connection is limited by the congestion window, rather than by the socket buffers. Finally, we present an application-layer mechanism (SOBAS) that automatically adjusts the socket buffer size close to its optimal value, based on direct measurements of the maximum received throughput and of the round-trip time, without requiring prior knowledge of the path characteristics.