In this paper, we put forward a HOListic traffic control framework for datacenter NETworks (HOLNET). HOLNET reformulates the network utility maximization (NUM) framework into a HOLNET NUM framework that fully harnesses the potential of the existing NUM-based solutions to allow large families of traffic control protocols of various degrees of sophistication to be developed, i.e., host-based, single or multiple Class-of-Service (CoS) enabled, single or multi-path congestion control, with or without in-network load balancing. Unlike the existing solutions that are largely empirical and point by design, HOLNET is a principled, systematic framework. All the protocols in a family developed under HOLNET share a common, user-defined global optimization objective and fairness criterion. As a result, the protocols in a family can be fairly compared and carefully selected to fully explore the performance, scalability and design complexity tradeoffs. Case studies, based on both a single and a multi-path host-based solutions, demonstrate the viability and flexibility in HOLNET design space exploration. To further test the backward compatibility and performance with respect to some existing lightweight solutions, we develop HOLNET-UTA, an integrated congestion control and load balancing protocol, achieving TCP-fair resource allocation. HOLNET-UTA is found by simulation to improve the average flow completion time (FCT) by more than 20%, compared to DRILL with DCTCP.

Developing electrophysiological recordings of brain neuronal activity and their analysis provide a basis for exploring the structure of brain function and nervous system investigation. The recorded signals are typically a combination of action potentials (spikes) and noise. High amounts of background noise and possibility of electric signaling recording from several neurons adjacent to the recording site have led scientists to develop neuronal signal processing tools such as spike sorting to facilitate brain data analysis. Spike sorting plays a pivotal role in understanding the electrophysiological activity of neuronal networks. This process prepares recorded data for meaningful interpretations of interactions among neurons, investigating the connectivity between different brain regions, and ultimately understanding the overall structure of brain functions. Spike sorting consists of three steps: spike detection, feature extraction, and spike clustering. There are several methods to implement each of spike sorting steps. This paper provides a systematic comparison of various spike sorting sub-techniques applied to real extracellularly recorded data from a rat brain basolateral amygdala. An efficient sorted data resulted from careful choice of spike sorting sub-methods leads to better interpretation of the brain structures connectivity under different conditions, which is a very sensitive concept in diagnosis and treatment of neurological disorders. Here, spike detection is performed by appropriate choice of threshold level via three different approaches. Feature extraction is done through PCA and Kernel PCA methods, where Kernel PCA outperforms. We have applied four different algorithms for spike clustering including K-means, Fuzzy C-means, Bayesian and Fuzzy maximum likelihood estimation. As one requirement of most clustering algorithms, optimal number of clusters is achieved through validity indices for each method. Finally, the sorting results are evaluated using inter-spike interval histograms. Rat brain real data used in simulations was recorded in neuroscience research center, Shahid Beheshti University of Medical Sciences, Tehran, Iran.

This paper proposes an algorithm for optimal decentralized traffic engineering in communication networks. We aim at distributing the traffic among the available routes such that the network utility is maximized. In some practical applications, modeling network utility using non-concave functions is of particular interest, e.g., video streaming. Therefore, we tackle the problem of optimizing a generalized class of non-concave utility functions. The approach used to solve the resulting non-convex network utility maximization (NUM) problem relies on designing a sequence of convex relaxations whose solutions converge to that of the original problem. A distributed algorithm is proposed for the solution of the convex relaxation. Each user independently controls its traffic in a way that drives the overall network traffic allocation to an optimal operating point subject to network capacity constraints. All computations required by the algorithm are performed independently and locally at each user using local information and minimal communication overhead. The only non-local information needed is binary feedback from congested links. The robustness of the algorithm is demonstrated, where the traffic is shown to be automatically rerouted in case of a link failure or having new users joining the network. Numerical simulation results are presented to validate our findings.

This paper describes a dynamical model-based method for the localization of road vehicles using terrain data from the vehicle's onboard sensors. Road data are encoded using linear dynamical models and then, during travel, the location is identified through continuous comparison of a bank of linear models. The approach presented has several advantages over previous methods described in the literature. First, it creates computationally efficient linear model map representations of the road data. Second, the use of linear models eliminates the need for metrics during the localization process. Third, the localization algorithm is a computationally efficient approach that can have a bounded localization distance in the absence of noise, given certain uniqueness assumptions on the data. Fourth, encoding road data using linear models has the potential to compress the data, while retaining the sensory information. Finally, performing only linear operations on observed noisy data simplifies the creation of noise mitigation algorithms.

This paper puts forward an end-to-end traffic control solution, which we refer to as TCP-Elastic Real-time SErvice (TERSE). TERSE provides a unified end-to-end traffic control protocol that enables Non-Real-time Elastic (NRE) service (i.e., the same as the one under the TCP control), Real-time Delay Adaptive (RDA) service, and Real-time Rate Adaptive (RRA) service. A specific service is enabled by properly setting a single parameter in the protocol. TERSE is underpinned by a sound design methodology. While having its roots in a well-known utility-based optimization approach, this methodology successfully addresses its limitations. It leads to a unified traffic control protocol, which has several provable properties, including fairness, convergence, and stability. The protocol is implemented in LINUX-based systems. The cross-Pacific testing of this protocol shows that it can achieve more than 1.2Mbps throughput performance, 150% higher than TCP-reno and 50% higher than TCP cubic. Both analytical and simulation studies also show that it can provide soft minimum-rate guarantees for both RRA and RDA traffic flows. Moreover, simulation also demonstrates that it is resilient to network resource shortage. As part of the protocol design, an effective utility function of TCP and the corresponding control law are derived, which captures TCP behavior not only in a qualitative but also in a quantitative manner.

This paper addresses the problem of segmenting a combination of linear subspaces and quadratic surfaces from sample data points corrupted by (not necessarily small) noise. Our main result shows that this problem can be reduced to minimizing the rank of a matrix whose entries are affine in the optimization variables, subject to a convex constraint imposing that these variables are the moments of an (unknown) probability distribution function with finite support. Exploiting the linear matrix inequality based characterization of the moments problem and appealing to well known convex relaxations of rank leads to an overall semi-definite optimization problem. We apply our method to problems such as simultaneous 2D motion segmentation and motion segmentation from two perspective views and illustrate that our formulation substantially reduces the noise sensitivity of existing approaches.

This work addresses the problem of optimal traffic engineering in a connectionless autonomous system. Based on nonlinear control theory, the approach taken in This work provides a family of optimal adaptation laws. These laws enable each node in the network to independently distribute traffic among any given set of next hops in an optimal way, as measured by a given global utility function of a general form. This optimal traffic distribution is achieved with minimum information exchange between neighboring nodes. Furthermore, this approach not only allows for optimal multiple forwarding paths but also enables multiple classes of service, e.g., classes of service defined in the differentiated services architecture. Moreover, the proposed decentralized control scheme enables optimal traffic redistribution in the case of link failures. Suboptimal control laws are also presented in an effort to reduce the computational burden imposed on the nodes of the network. Finally, an implementation of these laws with currently available technology is discussed.

In this paper, we address the problem of optimal decentralized traffic engineering when multiple paths are available for each call. More precisely, given a set of possible paths for each call, we aim at distributing the traffic among the available paths in order to maximize a given utility function. To solve this problem, we propose a large family of decentralized sending rate control laws having the property that each of the members of this family "steers" the traffic allocation to an optimal operation point. The approach taken relies on the control theory concept of Sliding Modes. These control laws allow each ingress node to independently adjust its traffic sending rates and/or redistribute its sending rates among multiple paths. The only nonlocal information needed is binary feedback from each congested node in the path. The control laws presented are applicable to a large class of utility functions, namely, utility functions that can be expressed as the sum of concave functions of the sending rates. We show that the technique can be applied not only to usual rate adaptive traffic with multiple paths, but also to rate adaptive traffic with minimum service requirements and/or maximum allowed sending rate and to assured service with targeted rate guarantee, all allowing for multiple paths. It is also shown that these control laws are robust with respect to failures; i.e., they automatically reroute traffic if a link failure occurs. Finally, we provide some insight on how to choose the "right" control law. In particular, we provide a way of choosing a member of the family of control laws that reduces the sending rate oscillation caused by implementation constraints like delays and quantization. An example of application of the approach delineated in this paper is also presented. This example provides some insights on the implementation aspects and illustrates the robustness of the control laws developed in this paper.