Today’s data centers act as the primary infrastructure for emerging technologies. QoS imposes requirements for more attentive techniques that can deal with different characteristics of traffic classes and patterns. In this context, network flows can be classified into large and long-lived flows called elephant flows and mice flows, which are small and short-lived flows. According to the characteristics of the emerging technologies, e.g., IoT and Big Data, mice flows are dominant; Hence, it is crucial to improve Flow Completion Time (FCT) for such delay-sensitive flows. This paper presents Sieve, a new distributed Software Defined Networks (SDN) based framework. Sieve initially schedules a portion of the flows based on the available bandwidth despite their classes. We propose a distributed sampling technique which sends a portion of the packets to the controller. Furthermore, Sieve polls the edge switches periodically to get the network information rather than polls all switches in the network, and it reschedules elephant flows only. Mininet emulator and mathematical analysis have been employed to validate the proposed solution in 4-ary Fat-Tree DCN. Sieve provides less FCT up to around 58% for mice flows and maintains throughput of elephant flows compared to Equal Cost MultiPath (ECMP) and Hedera.

As micro traffic is dominant in data center networks, many conditions must be satisfied to guarantee a very short delay. In this context, the flexibility provided by software defined networks (SDN) represents an opportunity for proposing new quality of service (QoS) based solutions. In this paper, we propose a new SDN-based framework to utilize QoS mechanisms in providing better conditions for data center networks to deliver micro flows efficiently. We present a Micro-flow framework which deals with queue length as a sign to avoid congestion events so that the flow completion time of micro flows can be improved by adapting the data rate of background flows without any intervention of source end hosts for rapid reaction. By creating queues match the network topology, our framework probes the length of queues dedicated for micro flows such that the data rate of background flows on upstream switches can be updated proportionally to the length of the micro queue on the downstream switch. We evaluated our solution by Mininet in different scenarios. The experiments demonstrate that Cubic and New Reno with our framework give better completion time for micro flows compared to that when DropTail is used by Cubic and New Reno, mitigate the probability of congestion events and do not severely throttle the data rate of background flows.

The emergence of large data centers and virtualization needs better and smarter solutions for traffic scheduling and load balancing. Data centers benefit from SDN regarding centralized monitoring and management for traffic routing. In general, the traffic in the data center environment can be classified as elephant and mice flow. Researchers showed that there is a significant amount of data carried over elephant flows; therefore, it should be conserved and maintained thoroughly. In this work, we introduce a stochastic performance evaluation model for estimating blocked rate prediction and risk analysis of the elephant flows for a load balancing data center with fat-tree topology using the SDN paradigm. The general procedure of the evaluation includes the estimation of the distribution of the path available bandwidth, including bandwidth error tolerance. The proposed model relies on Monte Carlo simulation to generate future prediction behavior of the load balancing technique. The achieved results examined with Value at Risk (VaR) along with statistics to percept the complete picture of the load balancing behavior.

The performance problems of recent transport protocols applied in fixed and high-speed wireless networks call for research for an effective congestion control approach, which is able to work efficiently in both domains. In this paper we propose a framework incorporating a novel Multi-Domain congestion control with a novel Lightweight Performance Enhancing Proxy (PEP), which requires no client modification. The method was implemented in the Linux kernel and the Lightweight PEP in ns-3 and we carried out a comprehensive performance evaluation of our proposed framework. We demonstrate that our approach can achieve a robust behavior in the presence of wireless loss while keeping the TCP fairness in the wired domain. Moreover, we also present that our method provides a faster adaptation in the wireless domain compared to the currently used CUBIC TCP.

QUIC is a new transport protocol for transferring web traffic proposed by Google employing CUBIC as the default congestion control algorithm with a small but important modification where it uses a smaller congestion window reduction than the one used in the original CUBIC. Motivated by this change, in this paper we address the robustness and sensitivity of the multiplication decrease factor. More specifically, we investigate the page load time of QUIC as an impact of the multiplication decrease factor in different network environments by changing the loss rate, delay, buffer size, web page size, and bandwidth. We have found that the region where the impact of the change of the multiplication decrease factor in QUIC is pronounced when large web pages are downloaded with low round-trip time achieving 22% decrease in the page load time.

HTTP has been the protocol for transferring web traffic over the Internet since the 90s. However, over the past 20 years websites have evolved so much that today this protocol does not provide optimal delivery over the Internet and became a bottleneck in decreasing page load times. Google is pioneering in finding better solutions for downloading web pages and they implemented two new protocols: SPDY in 2009 and QUIC in 2013. Since the wide range deployment of SPDY clients and servers it has been revealed that in some scenarios SPDY can negatively affect the page transfer time mainly due to the fact that the protocol is working over TCP. To tackle these obstacles QUIC uses its own congestion control and based on UDP in the transport layer. Since QUIC is a very recent protocol, this paper could help further understand its operation and performance in a wide range of network scenarios. We present a comprehensive study about the performance of QUIC, SPDY and HTTP particularly about how they affect page load time. We found that none of these protocols is clearly better than the other two and the actual network conditions determine which protocol performs the best.

The volume of Internet traffic has been growing exponentially in the last years, and this trend is expected to intensify in the future. In typical backbone networks hundreds of thousands of flows, originated from different users and versatile applications, compete for the available resources. Such a heterogeneous mixture of traffic flows leads to a continuously changing environment, hence it is crucial to deeply understand the behavior of the underlying data transfer mechanisms regarding many features like stability, convergence and responsiveness. In this paper we study the characteristics of the recently proposed digital fountain based transport in dynamic network conditions, and carry out a comparison with the traditional approach relying on TCP's congestion control.

In the recent years, a significant research effort has been devoted to the development of bandwidth estimation techniques and tools due to the broad range of possible applications. The vast majority of bandwidth estimation algorithms are designed and optimized for wired networks. Therefore, these solutions not only provide inaccurate results in wireless environments but also rely on some information usually not known in advance or produce a severe additional load on the network. Specifically, in mobile networks the continuously varying characteristics of radio links make it an extreme challenge to estimate the currently available bandwidth. In this paper we present a bandwidth estimation method worked out for mobile networks, which models the dynamics of the bottleneck queue and identifies its busy periods. Our algorithm can estimate the unused bandwidth by exploiting the user-generated downlink network traffic with negligible extra load. The operation of the algorithm is demonstrated on real traffic traces captured by a mobile device in a 3G network.

We present the versatile functionality of our novel user behavior based traffic emulation system in this paper. We show the unique feature of the system, i. e., it is capable of working on different platforms (Windows, Android), on different access technologies (wired, WiFi, 3G) and as a remote controlled system on different sites (Europe, Asia, South America). Our examples exhibit some of the manifold traffic analysis possibilities as a result of this key functionality. We have also made our system available to the public [1].

In this paper a three-dimensional (size, duration and rate) flow traffic characterization framework is proposed. Detailed investigation of application clusters in the characterization space is discussed and examples from specific applications (P2P, gaming, social networking, video playback) are presented. The framework provides an in-depth understanding of Internet traffic as demonstrated on measured traffic from a commercial network.

In this paper a novel approach to provide satisfactory multiplayer gaming quality in mobile environment is presented. The paper has two contributions: (i) evaluation of the delay characteristics of multiplayer games in mobile environment based on extensive measurements to verify whether HSDPA access can provide a satisfactory gaming quality; (ii) improving the gaming quality with new server update time optimization algorithms taking advantage of the statistical delay characteristics.

Game traffic depends on two main factors, the game protocol and the gamers' behavior. Based on a few popular real-time multiplayer games, this paper investigates the latter factor showing how a set of typical game phases-e.g., player movement, changes in environment- impacts traffic. By understanding the nature of this impact an algorithm is introduced to grab specific events and states from passive traffic measurements. Further, a measurement example including a detailed analysis is shown from an operational broadband network. The collected data of player behavior in the gaming environment was analyzed and it was also shown that the long range dependence (LRD) property of gaming traffic is not due to the popular explanation based on the heavy-tailed periods of player activities.

The introduction of High Speed Downlink Packet Access (HSDPA) presents new challenges to be solved in the UMTS Terrestrial Radio Access Network. Bandwidth reservation for HSDPA is not efficient and TCP cannot efficiently resolve a congestion situation because lower layer retransmissions hide the congestion situations from TCP. An HSDPA Flow Control algorithm was introduced by 3GPP to control congestion. This HSDPA Flow Control algorithm was originally intended to control radio scheduler queues in Node B. In this paper we propose an algorithm that also provides congestion control in the transport network. The performance analysis concentrates on transport network limited scenarios and shows that the algorithm can achieve high end-user perceived throughput, while maintaining low delay and loss in the transport network. The analysis also shows the advantages of the newly introduced congestion detection functionality.

Skype uses strong encryption to secure communications inside the whole Skype network. Clients choose communication ports randomly. Therefore traditional port based or payload based identification of Skype traffic is not feasible. In this paper we introduce a novel flow dynamics based identification method to discover Skype host and voice calls as well. Our method uses only packet headers and the extracted flow level information. The whole identification process is scripted in Transact-SQL, thus it can be executed automatically. We also present the validation of the algorithm together with some analysis of the identification results.

The paper is concerned with a novel adaptive game server protocol optimization to combat network latencies in the case of heterogeneous network environment. In this way, game playing becomes feasible for clients accessing the game via different networks, which can pave the way to securing a higher income for the game industry and service providers. The sever based game protocol is viewed as a number of arrival processes from the clients (which are characterized by different delays) and periodical updates sent by the server to the clients. Game quality is quantified by two measures: (i) the tail probability of the maximal idle time; and (ii) the probability of missing an update period. Our objective is to chose server update time subject to minimizing the probability of the maximal idle time exceeding a certain threshold (which threshold is associated with the "psycho-physically approved quality of the game"). In order to calculate the optimal server update time and the underlying tail distribution, statistical tools from large deviation theory are used. Furthermore, an adaptive on-line algorithm has been developed which can adjust the server update time by estimating the corresponding delay probability density functions based on past observations. Due to the new method, game quality can be significantly improved despite the wide range of client latencies which typically characterize the heterogeneous network environment. In this way, more players can be served which can further increase the business potential of network games. The performance of the new method has been evaluated by using measurements and extensive simulations.

In this paper we develop a simple but effective mathematical model to capture the file population dynamics of file-sharing peer-to-peer systems. Our modeling framework is based on the theory of branching processes. We describe analytically the behavior of the proposed model. The precise characterization of the necessary and sufficient conditions of population extinction or explosion is given based on the system parameters. We also present the expected ratio of active, passive and dead peers for the long-term regime. We validate and demonstrate our results in several simulation studies. Based on our results we propose a number of engineering guidelines to the design and control of file-sharing P2P systems.

The utilization of the free capacity by the TCP congestion control is not perfect due to the complex fluctuations of the background traffic resulting in reduced throughput. Different frequency components of the background traffic have different effects on the TCP performance. In this paper we present a comprehensive TCP performance evaluation study to understand the nature of the TCP adaptivity mechanism in frequency domain. Through simulations we analyze the impact of network parameters and also some alternatives for compensation of the throughput reduction. Finally, based also on the analysis of measured TCP traffic we investigate the robustness of TCP to fluctuations occur in actual network traffic.

In this paper, a state-based modelling of TCP traffic is presented. During a connection, TCP stays in one of the following states: slow start, congestion avoidance, loss recovery (fast recovery and/or fast retransmit) and time out. We consider the states of a TCP connection as the phases of a stochastic process. We propose the use of the discrete-time batch Markovian arrival process (D-BMAP) to model the traffic generated by a TCP connection. The main contributions of the paper are the following. Firstly, we provide a simple unified model for some well-known versions of TCP based on the D-BMAP process. Secondly, we introduce a new concept, namely the TCP characterization matrix for a TCP connection that characterizes the transition probabilities between the states of TCP. This matrix is crucial in our state-based analysis. Thirdly, we present a technique to detect the states of TCP. We have developed our technique into a tool called TCP-ASD that automates state detection of a TCP connection. Our tool can automatically detect the beginning and the end of the states of TCP and thus the sojourn time distributions as well as other statistics that we use in our analysis. We also discuss the tradeoffs between simplicity and accuracy in the state-based approach. Finally, we use simulation and numerical analysis to validate our proposed model.

We propose a new multifractal traffic model for network traffic. The model is a combination of a multiplicative cascade with an independent lognormal process. We show that the model has all the important properties observed in data traffic including long-range dependence (LRD), multifractality and lognormality. We also demonstrate that the model is flexible enough to capture the complete multifractal characteristics of data traffic including both the scaling function and the moment factor. On the other hand, we argue that the model is simple from practical point of view having only three parameters. Practical applications for measured data traffic and validation of the model with queueing performance evaluation are also presented.

In this paper a new traffic control method called advanced round robin (ARR) is presented. It is shown that ARR can provide both worst case and statistical quality of service guarantees without coupling cell loss and delay requirements. A related connection admission control algorithm with performance study is also presented. A comparative performance analysis of ARR using simulation was performed in different real scenarios. Results are compared to a reference system and to the simple round robin method.

The fractal nature of Internet traffic has been observed by several measurements and statistical studies. A new monofractal stochastic process called Limit of the Integrated Superposition of Diffusion processes with Linear differential Generator (LISDLG) is presented, which effectively characterizes network traffic monofractality. Several properties of the LISDLG model are presented including the covariance structure, cumulants, spectrum and bispectrum. The model captures high-order statistics by means of the cumulants. The relevance and validation of the proposed model are demonstrated by application studies for measured Internet traffic.

We show that the reachability and the observability of discrete time, time varying linear systems is equivalent to a structured Kalman rank condition, under the difference algebraic independence of the time varying coefficients of the structure matrices. In the case, when the system is not reachable, we only can state that the dimension of the reachability subspace reaches its maximum at infinitely many times. The observability of a bilinear system is also considered.

Measured traffic traces from live packet networks often contain non-stationary effects like level shifts or polynomial trends. In these cases the popular tests for long range dependence (LRD) can result in wrong conclusions and unreliable estimation of the Hurst (1950) parameter. We investigate the implications of these effects on several LRD tests. The use of these results can be utilized to avoid pitfalls in LRD traffic modeling. Our results are supported by both analytical and simulation studies with examples taken from traffic analysis of a live ATM network.

This paper presents a general traffic control framework for Asynchronous Transfer Mode (ATM) networks with its performance evaluation. The proposed traffic control scheme can incorporate all the recently considered ATM service classes including Constant Bit Rate (CBR), real time Variable Bit Rate (rtVBR), non-real time Variable Bit Rate (nrVBR), Available Bit Rate (ABR) and Unspecified Bit Rate (UBR) services. The control is based on a complete buffer partitioning architecture and on the associated buffer scheduling rule with adaptive weighting functions. We present, the formulation of the traffic control as an optimization problem in a 3-dimensional Quality of Service (QoS) state space. A solution approach based on dynamic programming is also suggested. A comprehensive performance evaluation of the method has been performed based on simulations and results are presented with several examples. The QoS dependence on CBR load, VBR load, VBR burstiness, UBR load are investigated and results are demonstrated with explanations.

TCP traffic round-trip delay measurements have been carried out in the Ericsson Corporate Network. Based on these data a comprehensive analysis has been performed to identify basic state changes in the network and to find the parametric distribution of the round-trip delay. The main finding of our research is that the round-trip delay can be well approximated by a truncated normal distribution. A method for parameter setting with the statistical tests for validation is also presented. Finally, we discuss the possible explanations of this result.