A Robustness Approach to Inter-Autonomous System Outbound Traffic Engineering

Many traffic engineering and network design tasks require a network-wide description as an input for their performance evaluation. Therefore, a traffic matrix represents the demands for these tasks and is applicable to them. Traffic matrix can be measured directly or estimated from an indirect data. Generating synthetic traffic matrix method, is a better way to achieve traffic matrix indirectly. We have previously proposed the conqueror traffic based synthetic traffic matrix and we have indicated that our conqueror traffic based matrix performed better in all cases where conqueror traffic existed. In this paper by measuring the maximum link-utilization, we consider how well the conqueror traffic based synthetic matrix optimizes the specific traffic engineering task; that is optimizing network routing to minimize congestion. First, we optimize the routing with the conqueror traffic based synthetic matrix to minimize congestion. Then, we test the performance of the resulting routing on the real traffic matrix. Simulation results indicate that OSPF optimization which we use in this paper to optimize the weights based on the conqueror traffic based synthetic matrix, optimize the routing better and result in a better performance on real traffic matrix.

Traffic matrices are required inputs for many IP network management tasks, such as capacity planning, traffic engineering, and network reliability analysis. However, it is difficult to measure these matrices directly in large operational IP networks, so there has been recent interest in inferring traffic matrices from link measurements and other more easily measured data. Typically, this inference problem is ill-posed, as it involves significantly more unknowns than data. Experience in many scientific and engineering fields has shown that it is essential to approach such ill-posed problems via regularization. This paper presents a new approach to traffic matrix estimation using a regularization based on entropy penalization. Our solution chooses the traffic matrix consistent with the measured data that is information-theoretically closest to a model in which source/destination pairs are stochastically independent. It applies to both point-to-point and point-to-multipoint traffic matrix estimation. We use fast algorithms based on modern convex optimization theory to solve for our traffic matrices. We evaluate our algorithm with real backbone traffic and routing data, and demonstrate that it is fast, accurate, robust, and flexible.

Traffic matrices are fundamental for network design, capacity expansion and traffic engineering. A Traffic matrix describes the rate or volume transferred between the ingress and egress nodes of a network. Internet traffic is dominantly TCP traffic and thus adapts to changing network conditions like routing or capacity. This effect is systematically neglected when using normal traffic matrices. This paper investigates the effect using three analytical models. It also shows how to use elastic traffic matrices for capacity expansion problems that do not neglect this effect.

Traffic Engineering (TE) is a widely-adopted network operation to optimize network performance and resource utilization. Destination-based routing is supported by legacy routers and more readily deployed than flow-based routing, where the forwarding entries could be frequently updated by TE to accommodate traffic dynamics. However, as the network size grows, destination-based TE could render high time complexity when generating and updating many forwarding entries, which may limit the responsiveness of TE and degrade network performance. In this paper, we propose a novel destination-based TE solution called FlexEntry, which leverages emerging Reinforcement Learning (RL) to reduce the time complexity and routing update overhead while achieving good network performance simultaneously. For each traffic matrix, FlexEntry only updates a few forwarding entries called <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">critical entries</i> for redistributing a small portion of the total traffic to improve network performance. These critical entries are intelligently selected by RL with traffic split ratios optimized by Linear Programming (LP). We find out that the combination of RL and LP is very effective. Our simulation results on six real-world network topologies show that FlexEntry reduces up to 99.3% entry updates on average and generalizes well to unseen traffic matrices with near-optimal load balancing performance.

Traffic engineering in hybrid SDN networks with multiple traffic matrices

Traffic matrices are essential for many traffic engineering tasks such as network management and capacity planning. However, it is difficult and cost to measure them accurately. The estimation of traffic matrix has usually been treated as a pure statistical inference problem. For example, infer the traffic matrix from the link load measurements and routing information. In practice, however, Internet Service Providers (ISPs) could measure partial traffic flows directly nowadays. In this paper, we propose a method to combine partial direct measurements of traffic matrix with one of the noticing method, the Minimum Mutual Information (MMI) method, to estimate the traffic matrix. Evaluation on real network data have demonstrated that with few levels of direct measurements of traffic flows could largely improve the performance of estimation, also the experiments have shown that choosing some large traffic flows to measure directly could have a great improvement on performance.

We consider the performance of estimated traffic matrices in traffic engineering. More precisely, we first optimize the routing in an IP backbone to minimize congestion with the estimated traffic matrix. We then test the performance of the resulting routing on the real traffic matrix.

Traffic matrix (TM) is a critical information for network operation and management, especially for traffic engineering (TE). Due to the technical and mercantile problems, real time measurement for TM is difficult in large scale networks. In this paper, we focus on predicting TM for dynamic traffic engineering. We propose several TM prediction methods based on Neural Networks (NN) and predict TM from three perspectives: predict the overall TM directly, predict each origin-destination (OD) flow separately and predict the overall TM combined with key element correction. In addition to the prediction accuracy, we evaluate different prediction methods through the performance of TE, as well as the prediction time. We test the proposed methods by real world datasets from Abilene, CERNET and GÉANT. The experiment results show that prediction methods based on Recurrent Neural Networks (RNN) can achieve better prediction accuracy than methods leveraging Convolutional Neural Networks (CNN) and Deep Belief Networks (DBN). Predicting each OD flow through RNN models can further improve the prediction accuracy, as well as the performance of TE under the OSPF network scenario, the SDN/OSPF hybrid network scenario and the multi-commodity flow problem scenario. However, it takes longer prediction time for predicting each OD flow sequence. In contrast, predicting the overall TM combined with key element correction can provide a trade-off between the TE results and the prediction overhead, which is more appropriate for dynamic TE in the above three scenarios.

In this paper we consider the problem of determining traffic matrices for end-to-end demands in an IP/MPLS network that supports multiple quality of service (QoS) classes. More precisely, we want to determine the set of traffic matrices Ti for each QoS class i separately. Ti contains average bandwidth levels for QoS class i for every pair of routers within the network. We propose a new method for obtaining QoS class specific traffic matrices that combines estimation and measurement methods: We take advantage of the fact that the total traffic matrix can be measured precisely in MPLS networks using either the LDP or RSVP-TE protocol. These measurements can then be used in a mathematical model to improve estimation methods - known as network tomography - that estimate QoS class specific traffic matrices from QoS class specific link utilizations. In addition to the mathematical model, we present results of the proposed method from its application in Deutsche Telekom's global IP/MPLS backbone network and we show that the estimation accuracy (mean relative error) is improved by a factor of 2.5 compared to results from network tomogravity. We investigate the structure of the estimated traffic matrices for the different QoS classes and motivate in this paper why QoS class specific traffic matrices will be essential for efficient network planning and network engineering in the future.

Traffic matrix describes the data flow between each pair of Origin-Destination (OD) over a measured period. However, it is very hard to be obtained in a large scale network. This paper compares two available diffusion operators. Based on the selection of good diffusion operator, we conduct multi-resolution analysis (MRA) on traffic matrices by diffusion wavelets. We also propose a method to detect the anomaly of the traffic matrix during a continuous period of time based on diffusion wavelet analysis.

It is very useful to infer traffic matrix (TM) from link measurements and routing information, especially for the tasks of capacity planning, traffic engineering and network reliability analysis. This inference problem is ill-posed as it involves more unknowns than data and the challenge lies in this problem is its ill-posed nature. To overcome this challenge, this paper firstly describes the inference problem into an optimization problem, in which the objective is to minimize the Euclidean distance between a certain predetermined prior and the target TM subject to the routing constraints upon the link measurements and TM. We secondly solve this problem by transforming the available routing matrix and applying the Lagrange multipliers method. We therefore propose a novel proposal on inferring TM, termed matrix partitioning and Lagrange multipliers (MPLM). An expression is derived for calculating the TMs from the link measurements and the transformations of the routing matrix. We analyze the computational complexity of MPLM, which comes out to be much less than the state-of-the-art approaches. We also suggest a new method to generate the prior. Numerical results are provided to demonstrate the accuracy of MPLM in estimating TMs, and the on-line TM estimation is also discussed in detail to show the techniques in MPLM.

An accurate traffic matrix (TM) is essential for network design, management and optimization. Software-defined networking (SDN) provides flow level statistics with global centralized control which enables construction of more accurate traffic matrices. However, retrieving all the flow statistics can cause a very significant overhead in the system. In this work, we propose an inference framework which utilizes Kalman filtering to create an accurate and timely traffic matrix. In our scheme, only a small number of flow statistics are measured at a time, yet the estimate of the TM is highly accurate. Besides, we propose a switch selection strategy which aims to minimize the entropy of the estimate, that is to maximize the information obtained by the estimation. Using simulation-based experiments, we show that our framework provides a very accurate TM estimate compared to the one constructed by using all the flow statistics in the network.

Traffic Engineering (TE) is important and necessary in order to fully utilize the existing network resources and reduce capital expenditure. Given its importance, many algorithms have been proposed in the literature. Unfortunately, there is still not a consistent methodology to evaluate these algorithms. Worse yet, some variants of the traffic engineering problem are known to be NP-hard. Thus, given a particular TE problem, in general, it is not possible to know the optimal solution; hence, it is difficult to assess how a particular heuristic algorithm performs. Even though several heuristic algorithms could be used to validate each other, the best solution from these algorithms could still be very far away from the optimal solution. We propose a novel methodology to evaluate TE algorithms. In this methodology, we construct TE problems with known optimal solutions and we then use these TE problem instances to test the performance of TE algorithms. We found that some TE algorithms perform poorly, and the result deviates from the optimum further as the problem size gets bigger. Our results suggest that there is large room for algorithm improvements and further research is required. Even though we only demonstrate the power of the methodology in the context of traffic engineering algorithms, the methodology is general enough that it could be applied in many other areas as well.

Existing traffic engineering (TE) solutions performs well for software defined network (SDN) in average cases. However, during peak hours, bursty traffic spikes are challenging to handle, because it is difficult to react in time and guarantee high performance even after failures with limited flow entries. Instead of leaving some capacity empty to guarantee no congestion happens due to traffic rerouting after failures or path updating after demand or topology changes, we decide to make full use of the network capacity to satisfy the demands for heavily-loaded peak hours. The TE system also needs to react to failures quickly and utilize the priority queue to guarantee the transmission of loss and delay sensitive traffic. We propose TED, a scalable TE system that can guarantee high throughput in peak hours. TED can quickly compute a group of maximum number of edge-disjoint paths for each ingress-egress switch pair. We design two methods to select paths under the flow entry limit. We then input the selected paths to our TE to minimize the maximum link utilization. In case of large traffic matrix making the maximum link utilization larger than 1, we input the utilization and the traffic matrix to the optimization of maximizing overall throughput under a new constrain. Thus we obtain a realistic traffic matrix, which has the maximum overall throughput and guarantees no traffic starvation for each switch pair. Experiments show that TED has much better performance for heavily-loaded SDN and has 10% higher probability to satisfy all (> 99.99%) the traffic after a single link failure for G-Scale topology than Smore under the same flow entry limit.

Current low earth orbit (LEO) satellite network display poor power efficiency, running network devices at full capacity all the time regardless of the traffic matrix and the distribution of the population over the Globe. Most of the research on energy efficiency of LEO satellites has focused on component level or link level. Therefore, this kind of research is not holistic to try to look at the satellite system as a single node. To enhance the energy efficiency. The solution should exploits multipath routing and load balancing. LEO network is overprovisioned, and hence selectively shutting down some satellite nodes and links during off-peaks hours seems like a good way to reduce energy consumption. In this paper, we exploit the fact that due to geographical and climatic conditions, some satellite links are expected to be loaded with data while others remain unused. Our approach is to power down satellite nodes and links during period of low traffic, while guaranteeing the connectivity and QoS. Finding the optimal solution is NP-problem and therefore, we explore in this work two heuristic algorithms. We evaluate our heuristics on a realistic LEO topology and real traffic matrices. Simulation results show that the power saving can be significant.