Y. Sinan Hanay

Saving energy and improving TCP throughput with rate adaptation in Ethernet

Reducing the power consumption of network interfaces contributes to lowering the overall power needs of the compute and communication infrastructure. Most modern Ethernet interfaces can operate at one of several data rates. In this paper, we present Queue Length Based Rate Adaptation (QLBRA), which can dynamically adapt the link rate for Ethernet interfaces at runtime using existing Ethernet standards. An implementation of the proposed rate adaptation functionality is demonstrated at runtime on a NetFPGA platform. Our results show that the rate adaptation approach can achieve significant energy savings and at the same time improve the throughput of TCP traffic due to the effect of packet pacing.

Formation control with potential functions and Newton's iteration

In this study, we analyze formation control (formation of a geometrical shape) of an autonomous multi-robot system with the use of artificial potential functions and Newtons iteration. The method is independent of the low-level vehicle dynamics of the robots and therefore it can be applied to different type of robots. We also perform numerical simulations to examine the performance of the method.

Network topology selection with multistate neural memories

We propose a novel algorithm to find efficient topologies for networks.Our method uses auto-associative neural memories.We implemented a simulator in C#, and we did simulations using realistic traffic.Our method improves a previously proposed method by 60%. We propose an efficient method utilizing neural memories for network topology selection. More specifically, we focus on virtual topology reconfiguration (VTR) problem in optical networks. One highly adaptive method that uses neural memories called Attractor Selection Based (ASB) algorithm was proposed before. However, ASB has an important drawback: it can work only with binary topologies since ASB uses binary neurons. In this work, our method is built on the same principles as ASB, yet it is capable of working with multistate topologies. By introducing multistate auto-associative memories, the information within a virtual network becomes finer grained than a binary state topology. The method we propose achieves a 60% performance improvement over ASB, and a 21% reduction in processing time in the simulations.

High-performance implementation of in-network traffic pacing

Optical packet switching networks promise to provide high-speed data communication and serve as the foundation of the future Internet. A key technological problem is the very small size of packet buffers that can be implemented in the optical domain. Existing protocols, for example the transmission control protocol, do not perform well in such small-buffer networks. To address this problem, we have proposed techniques for actively pacing traffic to ensure that traffic bursts are reduced or eliminated and thus do not cause packet losses in routers with small buffers. In this paper, we present the design and prototype of a hardware implementation of a packet pacing system based on the NetFPGA system. Our results show that traffic pacing can be implemented with few hardware resources and without reducing system throughput. Therefore, we believe traffic pacing can be deployed widely to improve the operation of current and future networks.

Virtual topology control with multistate neural associative memories

Previously, a highly adaptive virtual network topology (VNT) reconfiguration method called Attractor Selection Based (ASB) topology control was presented. ASB has an important drawback: it can only work with binary path tables. We propose a novel VNT controller by adding multistate path capabilities into ASB. However, adding multistate path capabilities reduces topology exploration space. We solve this problem by changing the system dynamics of ASB. With the modification of the system dynamics and the extension from binary to multistate paths, we observed a 60% performance improvement over ASB, and a 21% reduction in processing time in the simulations.

Sensor coverage maximization with potential fields

This paper proposes a novel algorithm to deploy mobile sensors for maximization of sensing area coverage. The algorithm presented is distributed and robust to errors. Our method is dynamic and based on potential fields. Although some methods using potential fields have been proposed previously, the approach we take here is different in constructing the potential field. In particular, we utilize an entropy-based potential function and use the Newtons iteration to determine the sensor positions. Simulations show that our method is more robust to errors, and less likely to be trapped in local minima compared to some previously proposed methods.

Improving resiliency against DDoS attacks by SDN and multipath orchestration of VNF services

We propose an architecture that increases the resiliency against DDoS attacks by leveraging virtual network functions (VNF) and software defined networking (SDN). In the first step, the proposed architecture places the virtual network functions (VNF) optimally by solving a linear program. In the second step, in order to add preemptive protection against DDoS attacks, special filter VNFs and secondary paths passing through these filter VNFs are set up by solving another linear program. Under a DDoS attack, SDN controller switches the routes affected by the attack to the secondary paths for filtering DDoS traffic in order to prevent over-utilization. The simulation results show that the proposed architecture can absorb higher amount of DDoS traffic with low impact on the average hop count.

Distributed sensor deployment using potential fields

Abstract Maximization of sensing coverage has been an important problem in mobile sensor networks. In this work, we present two novel algorithms for maximizing sensing coverage in 2D and 3D spaces. We evaluate our methods by comparing with two previously proposed methods. All the four methods are based on potential fields. The previous work used the same potential function, however the algorithms we propose here use two different potential functions. Potential fields require low complexity, which is crucial for resource lacking mobile sensor nodes. Though potential fields are widely used for path planning in robotics, only a few works used potential fields for coverage maximization in mobile sensor networks. Through simulations, we compare our proposal with the previous algorithms, and show that the algorithm we propose here outperforms previous algorithms.

Evaluation of topology optimization objectives

Two network-wide optimization contexts are traffic engineering and topology optimization. Various optimization objective functions and metrics have been proposed for both contexts. Yet, it is hard to evaluate the efficiency of those optimization objectives. Previously, a study analyzed the efficiency of some optimization metrics for traffic engineering by using linear programming (LP). On the other hand, in the topology optimization domain, there has not been any work on evaluation of different metrics. Because, it is hard to evaluate these metrics as the optimization algorithms are objective function tailored heuristics generally. As a result, a fair comparison of different objectives becomes hard. In this work, using machine learning we compare and analyze different traffic optimization objectives for topology optimization.

Topology selection criteria for a virtual topology controller based on neural memories

This work extends a previously proposed algorithm for virtual topology reconfiguration in all optical networks. Earlier, an algorithm using auto-associative neural memories has been presented. The algorithm stores topologies by assigning equal weights. In this work, we analyzed the effect of weighing topologies differently based on maximum flow, average weighted hop and the age of topology. Although we focus on optical networks, the algorithm and the analysis we present here can be useful in other network domains, such as wireless networks.