Version 8 (modified by Vic Thomas, 7 years ago) (diff)


Posters and Demonstrations at GENI NICE 2017


Steroid Openflow Service

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. A software defined networking based solution called Steroid OpenFlow Service (SOS) is a network service that transparently increases the throughput of data transfers across large networks. 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. In the demo we will show SoS running and webUI showing the real time plots of bandwidth utilization.


  • Khayam Anjam, Clemson University

Disk-to-Disk transfer using SOS

We have shown the increase in network performance using SOS in earlier demos. In this demo we will demonstrate the disk-to-disk transfer of large data sets using SOS and show the performance increase.


  • Junaid Zulfiqar, Clemson University
  • Khayam Gondal, Clemson University
  • Geddings Barrineau, Clemson University
  • Kuang-Ching Wang, Clemson University

On Balancing Load to Quickly Detect and Stop Attack Traffic

Our previous work proposed a control theoretic load balancer that offloaded traffic from an overloaded intrusion detection application (i.e., Snort) instance to another. We leveraged the management architecture of RINA to publish load and alert information from Snort instances to a Ryu SDN controller. In this demo, we generalize the framework with an “attack analyzer” that analyzes different kinds of intrusion alerts. On the GENI testbed, we generate DoS and port-scanning attack traffic using hping3 and Nmap tools, respectively. The controller communicates with the switch using OpenFlow to balance replicated traffic across Snort instances for analysis and to stop attack traffic. We show that under high load conditions, load balancing can help detect and stop attacks quickly. We show the impact of network delays and different control theoretic load balancers.


  • Nabeel Akhtar, Boston University
  • Marzieh Babaeianjelodar, Clarkson University
  • Ibrahim Matta, Boston University
  • Yaoqing Liu, Clarkson University

A Behavior-Driven Approach for Expressive Intent Specification in SDN and NFV

One of the goals of Software-Defined Networking (SDN) is to allow users to specify high-level policies into lower level network rules. Managing a network and decide what policies is appropriate requires, however, expertise and low level know-how. An emerging SDN paradigm is to allow higher-level network level decisions wishes in the form of intents"". Despite its importance in simplifying network management, intent specification is not yet standardized.

In this work, we propose an intent declaration approach based on Behavior-Driven Development (BDD). The level of expressiveness of our approach is maximal: intents are specified in plain English, and translated by our system into network policies, that are in turn, converted into low-level rules by the SDN (ryu) controller. Using the GENI testbed, in this work we demonstrate how to use our BDD framework to declare a few representative network intents: access control with a stateless firewall and traffic steering.


  • Flavio Esposito, St. Louis University
  • Thomas Merod, St. Louis University
  • Holly Wang, St. Louis University

Network Protocols

SPAN: Authentication protocol for software defined networking

SPAN is Multiparty Trust Negotiation (MTN) protocol that establishes mutual trust by the exchange of digital credentials and access control policies (ACP) among entities that may have no prior knowledge about each other. Research done in the area of automatic negotiation has been focus on creating an agreement between two parties however real world agreements involve more than two parties. In this paper we extend the stateless Eager Attribute Negotiation (SEAN) algorithm [9] to work in a multiparty environment. The proposed protocol is a distributed protocol and no centralized moderator is required. As a proof of concept we include an example that shows how this algorithm works.


  • Maha Allouzi, Kent State University
  • Javed Khan, Kent State University

Emergency High-Speed Internet Lane Protocol

The objective of emergency response is to minimize the impact of the event over time, particularly human casualties, environmental damage, and community disruption. Robust, reliable, and timely information sharing and dissemination is foundational to successful response. Thus, it is important to pursue information gathering to inform best practice response and rescue. Collecting and sharing the emergency data and information in a timely, reliable and effective manner to decision-makers, including Incident Command (IC) and to the responsible Public Safety Organizations (PSOs) through a regional center is vital to the success of emergency response. Current traffic routing in the Internet is subject to frequent route changes and high churn rates leading to delayed, looping, and lost packets. Lost and delayed packets can be highly detrimental to rescue operations. It is important to handle transportation of emergency information between the IC and Emergency Management Office (EMO)/PSO such that they are minimally impacted by routing instability and delays due to other traffic in the Internet. A new protocol called the Multi Node Label Routing (MNLR) Protocol has been developed to operate transparently to the Internet Protocol, providing a high-speed Internet lane for emergency information. It is designed with an immediate failover mechanism—meaning that if a link or node fails, it uses an alternate path right away, as soon as the failure is detected. The protocol has been implemented in GENI and uses a novel addressing scheme with labels auto assigned to nodes as they join the network. In this demo, we will show the capabilities of MNLR that will allow for auto-configuration of the nodes with multiple labels and the rapid recovery upon failures in the network, using a live video stream as an example application running on the network. The MNLR protocol failure detection and recovery operations will be directly compared with both Open Shortest Path First and Border Gateway Protocol running in their own networks.


  • Nirmala Shenoy, RIT
  • Erik Golen, RIT
  • Supriya Kharade, RIT
  • Shashank Rudroju, RIT

AGRA: AI-augmented geographic routing approach for IoT-based incident-supporting applications

Applications that cater to the needs of disaster incident response generate large amount of data and demand large computational resource access. Such datasets are usually collected in real-time at the incident scenes using different Internet of Things (IoT) devices. Hierarchical clouds, i.e., core and edge clouds, can help these applications’ real-time data orchestration challenges as well as with their IoT operations scalability, reliability and stability by overcoming infrastructure limitations at the ad-hoc wireless network edge. Routing is a crucial infrastructure management orchestration mechanism for such systems. Current geographic routing or greedy forwarding approaches designed for early wireless ad-hoc networks lack efficient solutions for disaster incident-supporting applications, given the high-speed and low-latency data delivery that edge cloud gateways impose. In this demo, we present a novel Artificial Intelligent (AI)-augmented geographic routing approach, that relies on an area knowledge obtained from the satellite imagery (available at the edge cloud) by applying deep learning. In particular, we propose a stateless greedy forwarding that uses such an environment learning to proactively avoid the local minimum problem by diverting traffic with an algorithm that emulates electrostatic repulsive forces. In our theoretical analysis, we show that our Greedy Forwarding achieves in the worst case a 3.291 path stretch approximation bound with respect to the shortest path, without assuming presence of symmetrical links or unit disk graphs. We establish the practicality of our approach in a real incident-supporting hierarchical cloud deployment to demonstrate improvement of application level throughput due to a reduced path stretch under severe node failures and high mobility challenges of disaster response scenarios.


  • Dmitrii Chemodanov, University of Missouri
  • Jon Patman, University of Missouri

Next Generation Applications

Next Generation Vehicle Network Applications

In the near future, autonomous vehicles are expected to become a part of our daily lives. Secure, stable, and speedy vehicle network communication becomes one of the most important features to support this. At Kettering University, our team developed a vehicle network testbed in GM Mobility Research Center, which is a 22-acre vehicle test track in our campus. The vehicle testbed incorporates two types of wireless networks: 4G-LTE and DSRC (802.11p). Specifically, 4G-LTE is used for V2X application, while DSRC is used for V2V and V2I safety application. This testbed will be utilized on two major projects: AutoDrive challenge and Smart Belt Coalition project.

SAE International and General Motors (GM) have partnered to headline sponsor AutoDrive Challenge, which is a three-year autonomous vehicle competition that will task students to develop and demonstrate a full autonomous driving passenger vehicle. The technical goal of the competition is to navigate an urban driving course in an automated driving mode as described by SAE Standard (J3016) level 4 definition by year three. As one of the eight participant team, our vehicle testbed will support Kettering Bulldog Bolt team to develop the best autonomous vehicle on Chevrolet Bolt EV.

The Smart Belt Coalition was formed in 2015 and is a strategic partnership comprised of twelve transportation agencies and academic institutions located throughout Michigan, Ohio, and Pennsylvania. As part of its strategic planning process, the Smart Belt Coalition agencies have mutually agreed to advance the development and deployment of a Work Zone Reservation and Traveler Information System (WZ System) as a top priority project for 2017 and 2018. We will use our testbed for work zone-related V2I applications development and testing, such as lane reduction warning or reduced speed warning applications.


  • Yunsheng Wang, Kettering University
  • John Geske, Kettering University

A distributed multi-loop networked system for wide area control of large power grid

We are going to demonstrate a prototype system that includes distributed Cloud, SDN, and distributed power grid control applications instantiated in ExoGeni testbed. This system aims to design three interactive control loops in controlling the compute, network and the physical systems. The demo would show that the SDN network connecting the distributed application running in the Cloud. The SDN control would change the paths based on active end-to-end latency measurement and the application would show improved performance in term of physical system stability with latency awareness.


  • Haoqi Ni
  • Mohamed Rahouti, University of Southern Florida
  • Aranya Chakrabortty
  • Kaiqi Xiong, University of Southern Florida
  • Yufeng Xin, RENCI

A Planet-scale distributed collaboration system

We will demonstrate the Ignite Distributed Collaborative Visualization System (IDCVS), a system which permits real-time interaction and visual collaboration around large data sets on thin devices for users distributed about the wide area. The IDCVS provides seamless interaction and immediate updates even under heavy load and when users are widely separated:. We will show two users, one on the show floor and another on the west coast of the US, collaborating around a very large data set with response times under 150 ms.


  • Rick McGeer, US Ignite
  • Glenn Ricart, US Ignite

A GENI based application to support Virtual Reality based training activities in orthopedic surgery

This demonstration involves highlighting a distributed approach to training orthopedic medical residents using Virtual Reality (VR) based simulation environments; this application exploits the capabilities of the Global Environment for Network Innovation (GENI)'s national test bed infrastructure. Our demonstration will show how expert surgeons in different hospitals can interact with medical trainees at other locations and teach them the fundamentals of orthopedic surgery. The high-definition multimedia streaming and haptic interfaces associated with the VR environment will enable trainees to remotely observe, participate and practice surgical techniques virtually from different locations (and also provides ‘on demand’ access to such medical educational and training resources). The virtual environments will enable students to learn the appropriate way of performing the LISS plating process which is an orthopedic surgical process to treat fractures of the femur. We are working with Dr. Miguel Pirela-Cruz at the Texas Tech Health Sciences Center (TTHSC) in El Paso, Texas.


  • J. Cecil, Oklahoma State University
  • Avinash Gupta, Oklahoma State University
  • Parmesh Ramanathan, University of Wisconsin-Madison
  • M. Pirela-Cruz, Texas Tech University

Clouds and Distributed Systems

Chameleon Stitching to ExoGENI

The demo will show the new Chameleon capability to stitch to ExoGENI and discuss other networking features soon to be available on Chameleon. In addition, the stitching demonstrate SAFE trust logic as mechanism for creating trust and security between slices residing on multiple testbeds and institutions.


  • Paul Ruth, RENCI

Extendable and Scalable IoT Middleware Through Multi-layer Virtual Sensor

Internet of Things (IoT) is an integral component of future Internet architecture where objects (i.e. sensors and actuators) are connected to each other via Internet to send and receive data. Objects are heterogeneous and communication protocols vary based on sensor type. Therefore, to enable applications to communicate with objects, a middleware is typically used to integrate objects and abstract the details of configurations. While this seems to be feasible, it involves many challenges of objects integration, protocol exchange, and data transfer and storage policies which require careful design patterns and solid implementation. In addition, applications are now using cloud resources and capabilities to be easier and more efficiently developed which adds more to the complexities of middleware design.

Therefore, in this demo, we will show a new multi-tiered middleware design that addresses IoT integration with cloud resources and provides applications with a good level of programming abstraction, scalable services, and efficient communication and protocol exchange. The middleware is featured with the principal of service-oriented and event-based design patterns and includes a new feature of multi-layered virtual sensor/actuator which simplifies data transfer and objects communication. We also use our testbed (SAVI) and its features to demonstrate our middleware capabilities and to efficiently test our agile development. The SAVI testbed includes many features of a modern software-defined infrastructure that leaves us with enough options to develop, test, and refine our product features.


  • Morteza Moghaddassian, University of Toronto
  • Hamzeh Khazaei, University of Toronto
  • Ali Tizghadam, University of Toronto
  • Hadi Bannazadeh, University of Toronto
  • Alberto Leon-Garcia, University of Toronto

Testbeds and Federation

Designing an Exploratory Testbed for Hyperprofile-based Computation Offloading

Recent offloading frameworks have been proposed with the goal of improving management decisions for offloading energy- and latency-sensitive tasks from mobile devices to nearby edge servers. Many of these solutions, however, are only evaluated via simulations, which may not accurately model real network behavior. We recently proposed a unique solution to the offloading problem which incorporates concepts from Knowledge-Defined Networking (KDN) to make intelligent predictions about offloading costs based on historical network data. This solution, known as hyperprofile-based computation offloading, represents each server instance as a node in a multidimensional feature space known as the hyperprofile. Each dimension of the hyperprofile corresponds to a feature that was predicted using a pre-trained machine learning model. Nodes within the hyperprofile are selected for offloading based on their proximity to a user-specified objective coordinate (e.g. the origin). A shortcoming of these data-driven approaches is that they rely on a representative training dataset and are very sensitive to training error. We created a synthetic dataset using NS-3 simulations and a real-world dataset from experiments in a GENI testbed in order to determine the viability of using synthetic data for accurately predicting real-world quantities. The data transfer characteristics of each data modality can be modeled by measuring transmission times for various sizes of data under varying network conditions. An analysis of using synthetic data for training and real data for testing is presented, as well as a comparative study between the two modalities. A discussion of the potential trade-offs and improvements for each data modality is also presented.


  • Jon Patman, University of Missouri
  • Flavio Esposito, St. Louis University
  • Prasad Calyam, University of Missouri

A Viral Planet-scale Infrastructure

We will demonstrate a PlanetLab-like system based on a modern, microservices-oriented platform. It features significant improvements over previous distributed systems infrastructures (PlanetLab, Seattle, and GENI): it is the first general-purpose system designed for rapid, viral expansion and deployment. In this demonstfration, we will add a node to the infrastructure in 15 minutes and deploy an application in 5,


  • Rick McGeer, US Ignite
  • Glenn Ricart, US Ignite

Elascale: Application Monitoring and Autoscaling as a Service in SAVI Testbed

Autoscalability is one of the crucial functionalities necessary for cloud software systems nowadays. Elascale strives to adjust both micro/macroservices' resources (using it's default autoscaling engine), with respect to workload and changes in the internal state of the whole application stack. We have implemented Elascale on SAVI Testbed: an instance will be leveraged to add auto-scalability to a generic IoT application. Furthermore, all of the components of Elascale are deployed as containers. In this demo, we showcase the deployment of the Elascale autoscaling system along with a simple, generic IoT application in SAVI Testbed. First, we will deploy the application that contains three main components: virtual sensors (data generators), stream processor (performs data processing) and database (storage). Next, we will deploy Elascale autoscaling engine to monitor the application. For our test, we scale the virtual sensors in order to stress the stream processor. Elascale will then automatically scale the stream processor to handle the workload. All of the information will be shown on a User Interface and on Kibana dashboard.


  • Rajsimman Ravichandiran, University of Toronto
  • Hamzeh Khazaei, University of Toronto
  • Thomas Lin, University of Toronto
  • Hadi Bannazadeh, University of Toronto
  • Alberto Leon-Garcia, University of Toronto

Education, Tools

The Popper Experimentation Protocol: Applying DevOps to the Evaluation of Computer Systems

Current approaches to scientific research require time-consuming activities that do not advance our scientific understanding. For example, cleaning data and writing code to attempt to reproduce previously published research. Can we find a better way to create and publish workflows, data, and models? The Popper Experimentation Protocol ( is a series of simple, easy-to-follow steps for implementing experiments using a DevOps approach.

Modern OSS development communities have created tools and practices (DevOps) to manage large codebases, allowing them to deal with high levels of complexity, not only in terms of code, but with the entire ecosystem that is needed in order to deliver changes to software in an agile, rapidly changing environment. Popper repurposes DevOps in the context of scientific explorations.

We will illustrate how to make use of the Popper command-line tool in order to re-run an existing experiment using geni-lib to configure infrastructure. Subsequently, we will show how to make use of Ansible and Docker, as well as to implement post-analysis of results using Jupyter notebooks. Additionally, we will show how Popper can generate files that can be used to connect a GitHub project (a ""Popperized"" repo) with TravisCI to continuously validate experiments.


  • Ivo Jimenez, University of California, Santa Cruz
  • Michael Sevilla
  • Noah Watkins
  • Jay Lofstead
  • Carlos Maltzahn
  • Kathryn Mohror
  • Andrea Arpaci-Dusseau
  • Remzi Arpaci-Dusseau

Virtual Computer Networks Lab

Jupyter is a widely used open-source tool based on the IPython implementation that allows users to share and run code in a browser. We will demonstrate the functionality of Jupyter for network testbed experimentation. We will present classroom assignments that can be instrumented in GENI through Jupyter.


  • Bhushan Suresh, University of Massachusetts at Amherst
  • Divyashri Bhat, University of Massachusetts at Amherst
  • Michael Zink, University of Massachusetts at Amherst