Search

This paper deals with congestion control in a software defined network (SDN) setting. Presently, explicit router schemes, such as Explicit Congestion Notification (ECN), work in conjunction with the TCP protocol to handle congestion in a distributed manner. With the emergence of SDN and centralized control, it is possible to leverage the global view of the network state to make better congestion control decisions. In this work, we explore the advantages of bringing in global information into distributed congestion control. We propose a framework where the controller with its global view of the network actively participates in the congestion control decisions of the end TCP hosts, by setting the ECN bits of IP packets appropriately. Our framework can be deployed very easily without any change to the end node TCPs or the SDN switches. We also show 30x improvement over the TCP Cubic variant and 1.7x improvement over TCP/RED in terms of flow completion times for one implementation of this framework, using the Mininet emulator.

This article presents a primer/overview of applications of Artificial Intelligence and Machine Learning (AI/ML) techniques to address problems in the domain of computer networking. In particular, the techniques have been used to support efficient and accurate traffic prediction, traffic classification, anomaly detection, network management, network security, network resource allocation and optimization, network scheduling algorithms, fault diagnosis and many more such applications. The article first summarizes some of the key networking concepts and a few representative machine learning techniques and algorithms. The article then presents details regarding the availability of data sets for networking applications and machine learning software and toolkits for processing these data sets. Highlights of some of the standards activities, pursued by ITU-T and ETSI, which are related to AI/ML for networking, are also presented. Finally, the article discusses a small set of representative networking problems where AI/ML techniques have been successfully applied., Comment: 38 pages

This report deals with the design of handover schemes for radio access networks (RAN) in 5G networks, using programmable data plane switches. In 5G networks, the NG-RAN architecture splits the Base Band Unit (BBU) into Central and Distributed Units (CU and DU). This structure has created a mid-haul Network, connecting CUs and DUs. The recent advancements in dataplane programmability can be used to enhance system performance. We show how P4 switches can be used to parse the packets between DU, CU, and Back Haul (Core Network) for potential system improvements. In particular, we consider the scenario of mobile handover. The proposed protocol is called SMARTHO, illustrating a smart handover. Programming Protocol-Independent Packet Processors (P4) is a programming language designed to support specification and programming the forwarding plane behavior of network switches/routers. In SMARTHO, we use P4 Switches to intervene in the handover process for fixed-path mobile users. A resource pre-allocation scheme that reserves resources before the UE reaches a future cell, is proposed. The solution is implemented using a P4-based switch introduced between the CU and the DU. The P4 switch is used to spoof the behavior of User Equipment (UE) and perform the resource allocation in advance. The proposed SMARTHO framework is implemented in the mininet emulation environment and in a reconfigurable hardware environment using NetFPGA-SUME boards. The emulation results show a handover response time improvement of 18% for a tandem of two HOs and 25% for a tandem of three HOs. For testbed implementation, we used NetFPGA-SUME boards as P4 switches. The handover time was measured to be approximately 50~milliseconds in the experiments conducted.

Service providers employ different transport technologies like PDH, SDH/SONET, OTN, DWDM, Ethernet, MPLS-TP etc. to support different types of traffic and service requirements. Dynamic service provisioning requires the use of on-line algorithms that automatically compute the path to be taken to satisfy the given service request. A typical transport network element supports adaptation of multiple technologies and multiple layers of those technologies to carry the input traffic. Further, transport networks are deployed such that they follow different topologies like linear, ring, mesh, protected linear, dual homing etc. in different layers. Path computation for service requests considering the above factors is the focus of this work, where a new mechanism for building an auxiliary graph which models each layer as a node within each network element and creates adaptation edges between them and also supports creation of special edges to represent different types of topologies is proposed. Logical links that represent multiplexing or adaptation are also created in the auxiliary graph. Initial weight assignment scheme for non-adaptation edges that consider both link distance and link capacity is proposed and three dynamic weight assignment functions that consider the current utilization of the links are proposed. Path computation algorithms considering adaptation and topologies are proposed over the auxiliary graph structure. The performance of the algorithms is evaluated and it is found that the weighted number of requests accepted is higher and the weighted capacity provisioned is lesser for one of the dynamic weight function and certain combination of values proposed as part of the weight assignment., Comment: PCATN paper

Data centers require high-performance network equipment that consume low power and support high bandwidth requirements. In this context, a combinatorial approach was proposed to design data center network (DCN) equipment from a library of components in \cite{infocom}. This library includes power splitter, wavelength multiplexers, reconfigurable add-drop multiplexers and optical amplifiers. When interconnecting optical components, it must be ensured that the resultant network supports specified target bit-error-rates (typically, at most $10^{-12}$). This paper reports experiment conducted on component interconnections and their computed bit-error-rates. From the experimental analysis, it was observed that the desired objective can be decided by considering a zeroth-order threshold for optical power at the receiver and before the amplifier. This paves way for the theoretical evaluation of several other such designs using this empirically derived model., Comment: 4 pages

This paper deals with cellular (e.g. LTE) networks that selectively offload the mobile data traffic onto WiFi (IEEE 802.11) networks to improve network performance. We propose the Seamless Internetwork Flow Mobility (SIFM) architecture that provides seamless flow-mobility support using concepts of Software Defined Networking (SDN). The SDN paradigm decouples the control and data plane, leading to a centralized network intelligence and state. The SIFM architecture utilizes this aspect of SDN and moves the mobility decisions to a centralized Flow Controller (FC). This provides a global network view while making mobility decisions and also reduces the complexity at the PGW. We implement and evaluate both basic PMIPv6 and the SIFM architectures by incorporating salient LTE and WiFi network features in the ns-3 simulator. Performance experiments validate that seamless mobility is achieved. Also, the SIFM architecture shows an improved network performance when compared to the base PMIPv6 architecture. A proof-of-concept prototype of the SIFM architecture has been implemented on an experimental testbed. The LTE network is emulated by integrating USRP B210x with the OpenLTE eNodeB and OpenLTE EPC. The WiFi network is emulated using hostapd and dnsmasq daemons running on Ubuntu 12.04. An off-the-shelf LG G2 mobile phone running Android 4.2.2 is used as the user equipment. We demonstrate seamless mobility between the LTE network and the WiFi network with the help of ICMP ping and a TCP chat application.