In recent years, data-intensive applications have been increasingly deployed on cloud systems. Such applications utilize significant compute, memory, and I/O resources to process large volumes of data. Optimizing the performance and cost-efficiency for such applications is a non-trivial problem. The problem becomes even more challenging with the increasing use of containers, which are popular due to their lower operational overheads and faster boot speed at the cost of weaker resource assurances for the hosted applications. In this paper, two containerized data-intensive applications with very different performance objectives and resource needs were studied on cloud servers with Docker containers running on Intel Xeon E5 and AMD EPYC Rome multi-core processors with a range of CPU, memory, and I/O configurations. Primary findings from our experiments include: 1) Allocating multiple cores to a compute-intensive application can improve performance, but only if the cores do not contend for the same caches, and the optimal core counts depend on the specific workload; 2) allocating more memory to a memory-intensive application than its deterministic data workload does not further improve performance; however, 3) having multiple such memory-intensive containers on the same server can lead to cache and memory bus contention leading to significant and volatile performance degradation. The comparative observations on Intel and AMD servers provided insights into trade-offs between larger numbers of distributed chiplets interconnected with higher speed buses (AMD) and larger numbers of centrally integrated cores and caches with lesser speed buses (Intel). For the two types of applications studied, the more distributed caches and faster data buses have benefited the deployment of larger numbers of containers.

Software Defined Networking (SDN) with defining characteristics, such as "separation of data and control plane" and "centralizing network control with decision making", has significantly simplified network management. However, active monitoring techniques used to dynamically measure network traffic introduce additional overheads in the network, while a passive approach lacks accuracy in terms of traffic measurement. As a result, various efforts have been devoted to designing per-flow based network measurement system to address both accuracy and overhead challenges. Existing measurement techniques lack a multi-objective network measurement mechanism to overcome various overheads, like communication cost, controller computation, and accuracy in a real-time environment. Therefore, this paper presents a novel and practical solution to enable accurate real-time traffic matrix for the traffic measurement system in SDN. The solution is proposed to measure fine-grained monitoring task with less controller communication and computational cost with high accuracy. The solution is based on two measurement designs, namely: fixed and elastic schemas. Our experiments demonstrate that both fixed and elastic schemas achieve significant overhead reduction without compromising on accuracy. (C) 2016 Elsevier Ltd. All rights reserved.

With the recent rise in cloud computing, applications are routinely accessing and interacting with data on remote resources. As data sizes become increasingly large, often combined with their locations being far from the applications, the well known impact of lower TCP throughput over large delay-bandwidth product paths becomes more significant to these applications. While myriads of solutions exist to alleviate the problem, they require specialized software at both the application host and the remote data server, making it hard to scale up to a large range of applications and execution environments. A software defined networking based solution called Steroid OpenFlow Service (SOS) has been proposed as a network service that transparently increases the throughput of data transfers across large networks. In this paper, the SOS architecture is refined to support data transfer at scale. In an OpenFlow-based cloud environment such as GENI, SOS can leverage the use of multiple agents to provide increased network throughput for many applications simultaneously. A cloud-based approach is particularly beneficial to applications in environments without access to high performance networks. This paper introduces the scalable SOS architecture and demonstrates its viability and scalability in GENI's distributed testbed.

Laparoscopic surgery is a minimally invasive surgical technique. The benefit of small incisions has a disadvantage of limited visualization of subsurface tissues. Image-guided surgery (IGS) uses pre-operative and intra-operative images to map subsurface structures. One particular laparoscopic system is the daVinci-si robotic surgical system. The video streams generate approximately 360 megabytes of data per second. Real-time processing this large stream of data on a bedside PC, single or dual node setup, has become challenging and a high-performance computing (HPC) environment may not always be available at the point of care. To process this data on remote HPC clusters at the typical 30 frames per second rate, it is required that each 11.9 MB video frame be processed by a server and returned within 1/30th of a second. We have implement and compared performance of compression, segmentation and registration algorithms on Clemson's Palmetto supercomputer using dual NVIDIA K40 GPUs per node. Our computing framework will also enable reliability using replication of computation. We will securely transfer the files to remote HPC clusters utilizing an OpenFlow-based network service, Steroid OpenFlow Service (SOS) that can increase performance of large data transfers over long-distance and high bandwidth networks. As a result, utilizing high-speed OpenFlow- based network to access computing clusters with GPUs will improve surgical procedures by providing real-time medical image processing and laparoscopic data.

Software defined networking (SDN) is lauded to be the paradigm of choice for the next generation networks. While academia explores use cases in various contexts, industry has focused on data centre networks' limited but intense needs. There is a significant range of complex and application-specific network services that can potentially benefit from SDN, but introduction and adoption of such solutions remains slow in production networks. One impeding factor is the lack of a simple yet expressive enough framework applicable to all SDN services across production network domains. Without a uniform framework, SDN developers create disjoint solutions, resulting in untenable management and maintenance overhead. In this paper, we propose an agent-based framework, which enabled the development of three distinct SDN services addressing the needs of network operators, application providers, and application users. The architecture facilitates application-oriented SDN design with an abstraction composed of software agents on top of the underlying network.

Mobile devices nowadays can find multiple wireless networks, such as WiFi, 4G/LTE and relay through devices. These networks have different characteristics in terms of coverage, data rate, and price. Meanwhile, mobile applications (and even different TCP/UDP connections) often have diverse and time-variant network needs. Thus, to better use all wireless network resources, it would be ideal to enable a TCP/UDP connection to 1) select the most appropriate network dynamically and 2) migrate between networks transparently. However, existing methods fail to provide both functions in a systematic and efficient way at the TCP/UDP connection level. In this paper, we adopt Software-Defined Networking (SDN) to realize such a feature. We use the features of SDN to realize intelligent network selection that is adaptive to time-variant application needs, network availability, and scheduling commands. To support transparent migration, an intelligent home agent (HA) is designed with the SDN to anchor packets from the mobile device. It can intelligently determine which wireless network a TCP/UDP connection is running over. Finally, our implementation demonstrates the effectiveness and efficiency of the proposed system.

The use of computer networks to relay streaming multimedia content is on the rise and is expected to continue to increase in the coming years. This paper discusses the use of GENI Cinema as a means to distribute streaming content across different networks at scale. Through the use of OpenFlow, GENI Cinema is able to gain control over the flow and routing of video streams in the network to efficiently and scalably distribute video content from video producers to video consumers. GENI Cinema accomplishes video "channel" switching with minimal perceived delay and no disruption in content to the producers or consumers. A prototype implementation of GENI Cinema on the distributed GENI testbed is discussed.

Leveraging multiple wireless technologies and radio access networks (RANs), vehicles on the move have the potential to get robust connectivity and continuous service. To support the demands of as many vehicles as possible, an efficient and fast network selection scheme is critically important to achieve high performance and efficiency. So far, prior works have primarily focused on design of optimization algorithms and utility functions for either user or network performance. Most such studies do not address the complexities involved in the acquisition of needed information and the execution of algorithms, making them unsuitable for practical implementations in vehicles. This paper proposes a fast cloud-based network selection scheme for vehicular networks. By leveraging a compute cloud's abundant computing and data storage resources, vehicles can leverage wider scope network information for decision-making. Vehicles select best access networks through a coalition formation game approach. A one-iteration fast convergence algorithm is proposed to achieve the final state of coalition structure in the game. Through extensive simulation, the proposed network selection scheme was shown to balance system throughput and fairness with a built-in utility division rule of the framework. The algorithm efficiency showed eightfold enhancement over a conventional coalition formation algorithm. Such features validate the potential of implementation in practice.

In the last decade, high-throughput DNA sequencing has become a disruptive technology and pushed the life sciences into a distributed ecosystem of sequence data producers and consumers. Given the power of genomics and declining sequencing costs, biology is an emerging “Big Data” discipline that will soon enter the exabyte data range when all subdisciplines are combined. These datasets must be transferred across commercial and research networks in creative ways since sending data without thought can have serious consequences on data processing time frames. Thus, it is imperative that biologists, bioinformaticians, and information technology engineers recalibrate data processing paradigms to fit this emerging reality. This review attempts to provide a snapshot of Big Data transfer across networks, which is often overlooked by many biologists. Specifically, we discuss four key areas: 1) data transfer networks, protocols, and applications; 2) data transfer security including encryption, access, firewalls, and the Science DMZ; 3) data flow control with software-defined networking; and 4) data storage, staging, archiving and access. A primary intention of this article is to orient the biologist in key aspects of the data transfer process in order to frame their genomics-oriented needs to enterprise IT professionals.

This paper details a framework that leverages Software Defined Networking (SDN) features to provide a testbed for evaluating handovers for IPv4 heterogeneous wireless networks. The framework is intended to be an extension to the Global Environment for Network Innovations (GENI) testbed, but the essence of the framework can be applied on any OpenFlow (OF) enabled network. Our goal is to enable researchers to evaluate vertical handover decision algorithms using GENI resources, open source software, and low cost commodity hardware. The framework eliminates the triangle routing problem experienced by other previous IPv4-compatible IP mobility solutions. This paper provides an overview of the testbed framework, implementation details for our installation using GENI WiMAX resources, and a discussion of future work.

This paper introduces GENI Cinema (GC), a system that provides a scalable live video streaming service based on dynamic traffic steering with software defined networking (SDN) and demand driven instantiation of video relay servers in NSF GENI's distributed cloud environments. While the service can be used to relay a multitude of video content, its initial objective is to support live video streaming of educational content such as lectures and seminars among university campuses. Users on any campus would bootstrap video upload or download via a public Web portal and, for scalability, have the video delivered seamlessly across the network over one or multiple paths selected and dynamically controlled by GC. The architecture aims to provide a framework for addressing several well-known limitations of video streaming in today's Internet, where little control is available for controlling forwarding paths of on demand live video streams. GC utilizes GENI's distributed cloud servers to host on-demand video servers/relays and its Open Flow SDN to achieve seamless video upload/download and optimization of forwarding paths in the network core. This paper presents the architecture and an early prototype of the basic GC framework, together with some initial performance measurement results.

This paper proposes a cloud-based architecture for enhancing the performance and capacity of vehicular networks of potentially multiple different wireless technologies. The approach addresses the well-known limitations of today's vehicle-initiated as well as base station-assisted handoff solutions; the former is reactive, therefore slow and inefficient, while the latter is mostly limited to within networks of a single technology. The handoff-as-a-service (HaaS) architecture leverages a cloud system's abundant computing and data storage resources to establish a database of key network properties and configuration options. By abstracting different networks' characteristics into a common set of descriptors, the database can aggregate and share properties of networks of different technologies. Leveraging network awareness of a wider scope, the HaaS service can further analyze optimal network configurations considering global efficiency and individual client requirements. The HaaS service in the computing cloud computes optimal handoff strategies on behalf of the vehicles, and OpenFlow is used to control both the vehicle and infrastructure side network interfaces seamlessly across multiple interfaces of different wireless technologies. This paper presents the proposed system architecture, its key components, and how they can be experimentally studied over National Science Foundation's Global Environment for Network Innovations (GENI) testbed. Experiment results on PC Engine device show the feasibility and advantage of the proposed handoff solution.