​GENI NICE Evening Demos

Location

Sheraton Fisherman's Wharf Hotel
2500 Mason St
San Francisco, CA 94133

Schedule

Tuesday Nov 10, 2015 - 5.30pm - 7.30pm

Session Leaders

Manu Gosain GENI Project Office	Peter Stickney GENI Project Office

Details

The evening demo session gives GENI NICE experimenters and developers a chance to share their work in a live network environment. Demonstrations run for the entire length of the session, with teams on hand to answer questions and collaborate. This page lists requested demonstrations categorized in broad interest groups. You can download project posters and supplemental information from attachments listed at the bottom of this page.

Projects

Education

Virtual Computer Networks Lab

Attendees interested in using GENI for education should attend this demo. In this demo we will present assignments that we have created within the scope of our GENI Virtual Computer Networks Lab. These assignments are executed on GENI testbeds and can be used by teachers in their Computer Networks or Distributed Systems classes. In addition to the assignments we will demonstrate LabWiki and how it can be used by students to execute the assignments. We will also demonstrate LabWiki’s features that support teachers in setting up and evaluating assignments.

Participants:

• Mike Zink, ​zink@ecs.umass.edu, University of Massachusetts
• Eswer Kishore
• Divyashri Bhat

Education Modules using GENI

This is a demo (with a poster at the demo station as well) of education modules that can be used in undergraduate networking classes. We have developed several curricular modules to teach basic networking concepts, including congestion control, TCP vs UDP, exploring router size, and several others.

If you teach networking courses, and would like toteands-on real demo, in-class tutorials, or assignments using GENI infrastructure to reinforce concepts through experimentation, please stop by our demo, and visit: geni.web.unc.edu

Participants:

• Jay Aikat, ​aikat@cs.unc.edu, The University of North Carolina at Chapel Hill
• Kevin Jeffay
• Ryan Doyle

Future Internet Architectures

Load Balancing Approach for Adaptive Bit-Rate Streaming in Information Centric Networks

The Information Centric Networking (ICN) paradigm promises deconstraining the current Internet architecture by allowing clients to directly address the desired content throughout the network. For the Internet, this is an evolutionary step from the idea of a narrow-waist core that only transports requests/replies to an intelligent architecture searching for and providing content. Multi-sourcing, which is one of the core ideas of ICN, constitutes a serious challenge for prevalent Internet applications such as video streaming. In this work we show how prominent adaptive video streaming protocols can benefit from the load balancing capabilities that are native to ICN. We examine the performance of content retrieval in ICN over Ethernet in a real-world testbed showing the impact of multi-sourcing and content size variation on the content transfer times.

Participants:

• Divyashri Bhat, ​dbhat@umass.edu, University of Massachusetts
• Cong Wang
• Amr Rizk
• Mike Zink, ​zink@ecs.umass.edu, University of Massachusetts

Efficient Caching for Dash

Dynamic Adaptive Streaming over HTTP (DASH) is a recent technology that delivers high quality streaming media content over the Internet using HTTP servers. A DASH server contains multiple representations for every video, with each representation varying in bitrate, resolution etc. Each of these representations are divided into smaller segments of fixed playback duration. the client downloads the individual segments during playback. The client employs an Adaptive Bitrate Algorithm (ABR) that determines the appropriate quality for the next segment based on the network conditions. The current Internet architecture utilizes in-network caching to improve content availability and performance. We identify the areas where DASH differs from other web requests, and propose an Adaptation Aware Cache (AAC) framework that optimizes serving DASH video streams by using bandwidth measurement at the cache and the knowledge of the adaptation scheme used at the client to predict the next segment request. This technique could be used to improve the cache performance by maximizing the byte-hits in cache and minimizing the unnecessary prefetches, thus improving the Quality of Experience (QoE) at the client-end.

To evaluate our proposal we used GENI testbed a web-server that host multiple DASH videos with several representations. All the segment request from the client are serviced using a single machine acting as a cache. The cache measures the current bandwidth based on one of the three throughput estimations: average of all values, average of last five values, and Harmonic average of all values.

Participants:

• Sheyda Kiani Mehr, ​skkv6@mail.umkc.edu, University of Missouri-Kansas City
• Parikshit Juluri
• Rohit Abhishek
• Deep Medhi

Network Protocols

A Flexible and Lightweight BGP Route Injector to Multiple Peers

The demo proposes and implements a flexible and lightweight BGP injector (mBGPInjector) that can announce both offline and real-time BGP routes to multiple peers with various filtering options. mBGPInjector also supports dynamical BGP configuration changes, such as online bringing a peer down or up, applying custom filtering rules and announcing or withdrawing arbitrary BGP routes. The implemented mBGPInjector is a lightweight perl program that consists of a few functions to read, analyze and direct BGP routes to corresponding connected BGP peers in a large size of network (eBGP or iBGP). Experiments with our mBGPInjector tool satisfy the requirements we need for research and education and demonstrate effectiveness under realistic workloads. In addition, we designed and tested the tool in Global Environment for Network Innovations (GENI) environment, where researchers and educators can take advantage of shared resource request files to repeat the same experiments without involvement of ponderous settings and configurations.

Participants:

• Yaoqing Liu, ​liu@clarkson.edu, Clarkson University
• Jialiang Pan

WiFi Multicast to Very Large Groups - Experimentation on the ORBIT Testbed

While WiFi has been proposed for multimedia content distribution, its lack of adequate support for multicast services hinders its ability to provide multimedia content distribution to a large number of devices. In our recent papers we proposed AMuSe, a scalable and adaptive system for WiFi multicast which is based on accurate receiver feedback and that incurs a small control overhead. Specifically, the system includes a scheme for dynamic selection of a subset of the multicast receivers as feedback nodes, which periodically send information, such as channel quality or received packet statistics, to the multicast sender.We implemented the AMuSe system in the ORBIT testbed and evaluated its performance in large groups with 150-200 receivers. We present a dynamic web-based application that demonstrates the operation of the system based on actual traces collected on the testbed in several experiments. It demonstrates the operation of AMuSe in various setting and environments.

Participants:

• Varun Gupta, ​vg2297@columbia.edu, Columbia University
• Raphael Norwitz
• Savvas Petridis
• Craig Gutterman
• Gil Zussman
• Yigal Bejerano

Next Generation Applications

A Cyber Physical Test Bed for Advanced Manufacturing

This demonstration will be a milestone in the area of Digital Manufacturing and involves showcasing a GENI based cyber physical framework for advanced manufacturing. This Next Internet based framework will enable globally distributed software and manufacturing resources to be accessed from different locations accomplish a complex set of life cycle activities including design analysis, assembly planning, and simulation. The advent of the Next Internet holds the promise of ushering in a new era in Information Centric engineering and digital manufacturing activities. The focus will be on the emerging domain of micro devices assembly, which involves the assembly of micron sized parts using automated micro assembly work cells.

Participants:

• J. Cecil
• Yajun Lu, ​yajun.lu@okstate.edu, Oklahoma State University
• Zak Zafar
• Esma Yahia

SDN & NFV

Network Measurement & Inference with SDN-enabled Online Learning

Fine grained information about the Internal Attributes of Interest (IAI) of a network, such as the per-flow size, delay, throughput or packet loss, provides an essential input for network design, capacity planning, routing protocol configuration and anomaly detection. In this poster, we would like to revisit the problem of network inference in the context of SDN-based networks. Using traffic matrix estimation (TME) as a case study, we propose a new measurement & inference framework with SDN-enabled online learning and show the performance of our framework for TM estimation and (hierarchical) heavy-hitter detection.

Participants:

• Chang Liu, ​cchliu@ucdavis.edu, University of California-Davis
• Mehdi Malboubi
• Chen-Nee Chuah

Virtual Network Migration Mechanism on GENI Platform

Network virtualization provides flexibility, enables agility and increases manageability by allowing coexistence of multiple virtual networks on the same physical substrate. Virtual network is built on top of the physical infrastructure and is assigned a subset of the underlying physical network resources. To have a better resource management, to recover from failure or provide defense against attacks, virtual networks may need to be remapped to different physical locations from time to time. However, there has not been a lot of work addressing the challenges of deploying a virtual migration mechanism in real infrastructure and exploring how the interaction between the virtual network and substrate network can affect the desired migration. In our project, we design and evaluate a virtual network migration mechanism in Openflow-enabled GENI platform. Specifically, we want to explore (1) how to deploy virtual network on GENI platform, (2) how to design a migration controller to make migration quick and automatic, and (3) how to minimize the disruption caused by migration. We will reveal the challenge and restriction to conduct virtual network migration experiments on GENI, and give recommendations for GENI platform to enhance their ability to support virtual network migration experiments.

Participants:

• Yimeng Zhao, ​zhaoym428@gmail.com
• Samantha Lo
• Niky Riga
• Mostafa Ammar
• Ellen Zegura

Emergency Response System

My demo is based generally on the openFlow architecture, and more specifically my demo consists of:

• Floodlight
• OVS Swithes(Open vSwitch): precisely I am using 10 of these Switches
• 2 nodes, 1 as the client and the other one is the server.

The purpose of my research is finding the best algorithm that guaranties the fastest communication between the client and the server node when n numbers of OVS Swithes are being interconnected to the client and the server host.

Then the next step of my research will be improving this algorithm to cover the security part, and how to avoid the communication attacks.

Participants:

• Mohamed Rahouti, ​mrahouti@mail.usf.edu, University of Southern Florida
• Dr. Kaiqi Xiong

Application Centric Network Orchestration Framework: ODENOS

We demonstrate a virtual network orchestrator framework, called ODENOS (Object-DEfined Network Operating System). ODENOS allows end users and application service providers to create and control virtual networks on top of heterogeneous (multi-layer and multi-domain) WAN (Wide Area Network) in an integrated way. ODENOS has a powerful graph-based network abstraction that can accommodate different kinds of network domains and layers, where any kind of network is abstracted as a graph of node, port, link, and flow. ODENOS also has a modularity and hierarchical structure, where modules can be stacked and are allowed to be inserted or removed. With the powerful abstraction, ODENOS can federate different network administrative domains. Also, ODENOS can slice a network and delegate the control of slices to application service providers. In addition, ODENOS can provide virtual network view and control API (Application Programming Interface) for each application service providers. ODENOS is an open source software released in February 2015, and distributed on GitHub. ODENOS is a part of Japanese government funded project, called O3, and the project member includes major Japanese telecom carriers and vendors, such as NTT, NTT Communications, Fujitsu, Hitachi and NEC. Our demonstration shows that ODENOS provides virtual networks that enable application service providers to satisfy requirements for application service quality without considering network architecture in detail. In the demonstration, ODENOS slices carrier networks consisting of multi-layer and multi-domain networks on demand, and performs application service provisioning smoothly. Types of application are video delivery and monitoring, and will be IoT (Internet of Things) service in the future. In addition, we show that ODENOS provides virtual network view of each service providers by user-friendly GUI (Graphical User Interface).

Participants:

• Satoshi Kamiya, ​kamiya@ak.jp.nec.com, NEC Corporation
• Kazuya Suzuki
• Masahiro Hayashitani
• Yoshiaki Kiriha

Building A Software Defined Networking-Based Communication Framework for Emergency Response

Timely and reliable data transfer between incident sites and emergency management office has played a key role in emergency response applications. In this project, we have been building a communication framework for ensuring minimal delay guarantee in emergency response networks by using Software Defined Networking (SDN) techniques. In this poster and demo, we will demonstrate the proposed communication framework whose experiments have been conducted on the Global Environment for Networking Innovations (GENI) testbed, and show the evaluation results. More precisely, we will demonstrate how to leverage an SDN controller to choose the best path between incident sites and emergency management office according to an objective, in this example, a path with minimal delay so as to minimize response delays in emergency communication networks.

Participants:

• Kaiqi Xiong, ​kqxiong18@gmail.com

GENI Cinema

Video streaming over the Internet, be it static or live streaming, is rapidly increasing in popularity. Many video streaming services exist to serve a variety of needs, such as video conferencing, entertainment, education, and the broadcast of live events. These services rely heavily on the server application to adapt to increasing and decreasing demand for a particular video resource. Furthermore, they require the reallocation of resources and the restart of the stream when a client stops, starts, and/or switches to a different stream. SDN and specifically OpenFlow can be creatively used to reallocate some of these tasks to the network and link layers.

Our goal is to provide a scalable service for GENI using OpenFlow that supports the broadcast of live video streams from an arbitrary number of video-producers to an arbitrary number of video-consumers, where video-consumers can change “channels” without disrupting their existing stream and without affecting the load on a particular video stream source.

Participants:

• Ryan Izard, ​rizard@g.clemson.edu, Clemson University
• Qing Wang
• Geddings Barrineau
• KC Wang

SDN/OpenFlow GENI lab exercises to measure and improve QoS/QoA/QoE

We have prepared a set of SDN/OpenFlow GENI lab exercises to measure and improve QoS/QoA/QoE for advanced Cloud Computing course and use one of these labs as a demo. During the demo we solve a problem of running a real research application called LOFT over a simple network infrastructure in a disaster scenario. The LOFT is used for regional scale tracking purposes and therefore requires fast image transferring from the collection site to the computation site. Moreover, in a disaster scenario regular networks usually are congested. To solve this real research problem, we perform traffic engineering in overlay network to overcome congestions. Finally, we show QoS, QoA and QoE improvement, based on following metrics: available Bandwidth (Objective QoS), Bandwidth Consumption (Objective QoA), File Transfer Time (Objective QoA), Mean Opinion Score (Subjective QoE). We note that this demo is based on one of the labs for advanced Cloud Computing course and can be used for education.

Participants:

• Prasad Calyam, ​calyamp@missouri.edu, University of Missouri
• Dmitrii Chemodanov

Managing NFV using SDN and control theory

This demo shows that control theory and SDN (Software Defined Networking) are key components for NFV (Network Function Virtualization) deployment. The management architecture of RINA (a clean-slate Recursive InterNetwork Architecture) is used to manage Virtual Network Function (VNF) instances over the GENI testbed.

In this demo, we will deploy an Intrusion Detection and Prevention System (IDPS) as the VNF. Our network topology has source and destination hosts, multiple IDPSes, an Open vSwitch (OVS) and an OpenFlow controller.

A distributed management application running on RINA measures the state of the VNF instances and communicates this information to the OpenFlow controller. The controller uses a control-theoretic approach to balance load across the VNF instances by updating flow rules on the OVS switch.

This demo demonstrates the benefits of RINA management and control-theoretic load balancing in virtualized environments. It also shows that GENI can easily support a wide range of SDN and NFV related experiments.

Participants:

• Ibrahim Matta
• Nabeel Akhtar, Nabeel Akhtar, Boston University

Experimental Demonstration of Brokered Orchestration for end-to-end Service Provisioning and Interoperability across Heterogeneous Multi-Operator (Multi-AS) Optical Networks

A broker on top of opaquely-managed optical domains advertising their capabilities is proposed to provision multi-AS connections in multi-operator scenarios. In case of no spectrum continuity, intra-domain spectral defragmentation is performed. Experimental assessment was conducted on a distributed multi-continental infrastructure.

Participants:

• Alberto Castro, ​albcastro@ucdavis.edu, University of California-Davis
• Lluis Gifre
• Cen Chen
• Jie Yin
• Zuqin Zhu
• Luis Velasco
• S. J. Ben Yoo

Architectural Issues in Virtualizing Intrusion Detection System as a Network Function

"Networking Services Providers face many challenges to introducing a new network service. Traditionally such services are offered through custom hardware appliances, which are difficult to deploy, have a limited life cycle, and are tied to a particular service. Virtualization of Network Functions promises many of the advantages that Cloud Computing has offered to traditional computing: efficient resource utilization, economies of scale, use of commodity hardware, elastic resource scaling, speedy deployment of new services, etc. We look at security features as network functions (NFs) that can be virtualized and offered as a service. In particular, we propose different ”security-as-a-service” architecture scenarios for intrusion detection/prevention system (IDS/IPS), and analyze the security and cost implications of the architecture choice. We create a framework to study the impact of architecture choices. We validate several of these architectures in a realistic deployment on GENI and also study their impact on network performance.

Our deployment on GENI uses dynamically reserved compute and network resources at multiple sites across the country to realistically emulate various cloud deployment scenarios. As future work, we plan on creating a number of virtualized network functions (VNFs) and orchestrating service function chains that can integrate in the ChoiceNet (NSF FIA) framework."

Participants:

• Harshvardhan P. Joshi, ​hpjoshi@ncsu.edu, North Carolina State University
• Rudra Dutta

A Public Safety 3D Surveillance Network

The project describes a mobile surveillance system to help law enforcement better perform public surveillance. Conventional and 3D cameras are mounted on the police vehicles to stream data in real-time to the cloud for processing. We use SDN to control the network flow to prioritize more important videos from less important ones. The 3D cameras are used to automatically detect potentially suspicious incidents in poor light conditions, e.g. person hiding in the dark, and acts as an input to the SDN controller. The poster will describe the overall surveillance system. The demo will include video clips of how the 3D cameras perform in the dark.

Participants:

• Md Zakirul Alam Bhuiyan
• Waqas Latif
• Pengpeng Liang
• Joshua Lloret
• Haibin Ling
• Chiu C. Tan, ​cctan@temple.edu, Temple University
• Jie Wu

Frequency-Minimal Moving Target Defense using Software Defined Networking

With the increase of cyber attacks such as DDoS, there is a need for intelligent counter-strategies to protect critical cloud-hosted applications. The challenge for the defense is to minimize the wastage of cloud resources and limit loss of availability, yet have effective proactive and reactive measures that can thwart attackers. In this poster, we address the defense needs by leveraging moving target defense protection within Software Defined Networking-enabled cloud infrastructure. Our novelty is in the frequency minimization and consequent location selection of target movement across heterogeneous virtual machines based on attack probability, which in turn minimizes cloud management overheads. We evaluate effectiveness of our scheme using a large-scale GENI testbed for a just-in-time news feed application setup. Our results show low attack success rate and higher performance of target application in comparison to the existing static moving target defense schemes that assume homogeneous virtual machines.

Participants:

• Saptarshi Debroy, ​debroysa@missouri.edu, University of Missouri
• Prasad Calyam

Symbiotic CAV Evolution: Software-Defined Infrastructure and Case Study in Public Safety

Connected and automated vehicles (CAVs) represent a paradigm shift in road-transportation and vehicle-centered experience. As a foundation for CAVs, vehicular wireless networking significantly impacts the feasibility, adoption, and performance of CAV applications. This inherent coupling between CAV networking and applications, the wide spectrum of CAV applications envisioned, and the continuous evolution of CAV networking and applications as well as the associated society impact require us to rethink our approaches to the research, development, and deployment of CAV networks and applications. Leveraging software-defined platforms and infrastructures for CAV networking and application deployment, we propose a promising approach that allows for symbiotic exploration and evolution of CAV applications and networks in shared real-world systems and environments. Using wireless-networked 3D vision for public safety as an example, we illustrate this approach using vehicular sensing, networking, and cloud computing resources in the national GENI (i.e., Global Environment for Network Innovations) infrastructure.

in particular, vehicles with software-defined CAV innovation platforms are deployed for supporting 3D vision for campus safety surveillance, vehicular sensing, and vehicular control networking emulation at the same time. Fusing real-time streamed videos with 3D campus environment, 3D vision facilitates campus safety surveillance; internal and external vehicle sensing enables modeling and development of CAV control algorithms for driving safety, which is also a major concern in public safety emergency response; integrating real-world vehicles with cloud-based simulation, vehicular control networking emulation enables exploration of next-generation vehicular wireless network solutions, which in turn enables future-generation CAV applications such as safe, green CAVs and real-time collaborative 3D vision for public safety emergency response.

Our work has been demonstrated to senior leadership in Washington, DC (e.g., leaders from the White House Office of Science and Technology Policy, NSF, NIST, and DHS) in March 2015, an a YouTube video about the demo can be found at ​https://www.youtube.com/watch?v=y_QxXA0MJzI.

Participants:

• Yuehua Wang
• Hai Jin
• Chuan Li
• Hongwei Zhang, ​hongwei@wayne.edu, Wayne State University
• Jing Hua

Security

Covert Storage Channel Analysis Using SDN and Experimentation on GENI

Covert channel communication is a method of sending information over a network without being noticed by normal users or network administrators and without alerting Intrusion Detection Systems (IDSs) or firewalls. Specifically, Covert Storage Channels (CSCs) use a common storage medium to send information secretly. Efficient detection of CSCs is critical to defeating data exfiltration and secret command and control networks. There are two major challenges in detecting and mitigating CSCs: (1) By its nature, it is difficult to differentiate CSC traffic from normal traffic. (2) Moreover, complex computer networks are generating big data that pose great challenges to perform CSC detection and mitigation timely and accurately. Even though there has been a wealth of work on detecting CSCs using stateful IDSs and traffic patterns, none of these works have employed the unique capabilities of Software Defined Networking (SDN). These capabilities rely on the separation of data from control plane to give a holistic picture on network flows. In this project we focus on a collaborative CSC detection and mitigation approach and its implementation. We employ SDN that offers a wealth of information about traffic flows, to improve both the efficiency and accuracy. We deploy monitors and correlators that dynamically coordinate with each other in CSC analysis. We study this approach for CSC analysis and mitigation on a realistic test bed, GENI (Global Environment for Network Innovations), and present preliminary results on the performance of this CSC detection approach.

Participants:

• Yiyuan Hu
• Xiangyang Li
• Xenia Mountrouidou, ​mountrouidoup@wofford.edu, Wofford College

DDoS Flooding Detection and Containment Using SDN and Experimentation on GENI

Distributed Denial of Service (DDoS) attacks are well studied in terms of detection and mitigation. However, as the network complexity and network traffic data has dramatically increased, these attacks have gained the spotlight again. Several cases of advanced persistent threats or other complex attacks are initiated with a DDoS flooding attack. Thus, it is important to detect and mitigate DDoS attacks accurately and efficiently. The problem is to distinguish DDoS from regular traffic and discover the instigator of the attack. Today’s increasing network complexities and the large volumes of data make this a challenging problem. We employ Software Defined Networking (SDN) that can enable targeted forensic evidence collection of computer network attacks. We present an innovative approach that coordinates distributed network traffic monitors that conduct anomaly detection and attack correlators that perform deep packet inspection supported by Open Virtual Switches (OVS). Our collaborative detection and mitigation approach looks for network flooding attack signature constituents that possess different characteristics in the level of information abstraction. Therefore, this approach is able to not only quickly raise an alert against potential threats, but also follow it up with careful verification to reduce false alarms. We experiment with this SDN-supported collaborative approach to detect DDoS attacks on the Global Environment for Network Innovations (GENI), a realistic virtual testbed. The response times and detection accuracy have demonstrated our method’s effectiveness and scalability.

Participants:

• Xenia Mountrouidou, ​mountrouidoup@wofford.edu, Wofford College
• Xiangyang Li
• Tommy Chin
• Kaiqi Xiong

Tools, Testbeds and Federation

CloudFinder: A System for Processing Big Data Workloads on Volunteered Federated Clouds

The proliferation of private clouds that are often underutilized and the tremendous computational potential of these clouds when combined has recently brought forth the idea of volunteer cloud computing (VCC), a computing model where cloud owners contribute underutilized computing and/or storage resources on their clouds to support the execution of applications of other members in the community. This model is particularly suitable to solve big data scientific problems. Scientists in data-intensive scientific fields increasingly recognize that sharing volunteered resources from several clouds is a cost-effective alternative to solve many complex, data- and/or compute-intensive science problems. Despite the promise of the idea of VCC, it still remains at the vision stage at best. Challenges include the heterogeneity and autonomy of member clouds, access control and security, complex inter-cloud virtual machine scheduling, etc. In this paper, we present CloudFinder, a system that supports the efficient execution of big data workloads on volunteered federated clouds (VFCs). Our evaluation of the system indicates that VFCs are a promising cost-effective approach to enable big data science.

Participants:

• Abdelmounaam Rezgui, ​rezgui@cs.nmt.edu, New Mexico Institute of Technology
• Nickolas Davis

Deploying Private ExoGENI Racks on CloudLab

The new mid-scale infrastructures deployed by NSFCloud are now available for use by the GENI community. These infrastructures are intended to be used for research into the future of cloud computing. This demonstration shows work toward enabling the deployment of private ExoGENI racks on CloudLab infrastructure. Private ExoGENI racks can be used as both a development space for the ExoGENI team to design and deploy new features at scale, as well as by GENI users who wish to perform controlled experiments on their own isolated ExoGENI rack.

The demonstration will include creating a CloudLab slice containing a functional ExoGENI rack, submitting a slice request to the ExoGENI rack using its native API. The ExoGENI slice will then be used to run the ADCIRC Storm Surge modeling workflow.

Currently, private ExoGENI racks can be deployed on the APT, Clemson, and Wisconsin CloudLab clusters. We have successfully deployed private racks containing as many as 64 nodes (512 cores). In turn, these private ExoGENI racks have been used to create slices containing as many as 512 VMs. In addition, these ExoGENI VMs can utilize advanced hardware capabilities, such as the SR-IOV enabled Infiniband network fabric available at the APT site. For example, an ExoGENI slice composed of 64 VMs (512 cores) has been used to run the ADCIRC storm surge model at near-native performance; fast enough to contribute to an actual urgent hurricane simulation.

Continued work will expand the capacities of private ExoGENI racks enabling them to connect to dynamic circuit providers. This capability will enable us to incorporate private ExoGENI racks residing in CloudLab into the ExoGENI federation, as well as enabling the deployment private ExoGENI-based federations composed of multiple private racks residing across many CloudLab sites.

Participants:

• Paul Ruth, ​pruth@renci.org, RENCI

FireAnt: An Autonomous and Distributed Framework for Scaling Heterogeneous Cloud Infrastructure

In a large-scale data center or multiple cloud clusters, control and management is a non-trival task. Control and management includes resource discovery, reservation, monitoring, maintenance, teardown, and etc. Centralized control of federation among different aggregate managers is a popular method recently, for example GENI deployment. However, such mechanism requires additional external infrastructure. This architecture is not able to scale infinitely due to the computing and access limitations of the control infrastructure. Furthermore, cloud infrastructure, e.g. OpenStack, itself does not address and solve this scalability issue when controlling thousands of nodes in a data center. Hence, FireAnt is proposed to solve this scalability issue, enabling extending from ten nodes to a million nodes.

In this demo, each cluster in the grid is an OpenStack cluster with FireAnt. A cluster without any free resources initiates a request and FireAnt sends it out for resource discovery. A remote cluster with sufficiently available resources accommodates this request and the local FireAnt calls OpenStack APIs to allocate resources locally (e.g. a VM). A reply is sent back to the origin cluster and VMs between these two clusters are stitched using VXLAN automatically by FireAnt. To end users, this process happens seamlessly and can be managed only in the origin cluster.

Participants:

• Ke Xu, ​ke_x@dell.com, Dell
• Rajesh Narayanan

Hybrid Cloud technologies for HPC queuing systems experiments for 'Simulation-as-a-Service'

Advanced manufacturing today requires diverse computation infrastructure for data processing. Our 'Simulation-as-a-Service' App, currently compute jobs over OSU HPC resources. However, there is a need to access to different computation resources available. We provide to users access to a variety of clouds such as Amazon and CloudLab as HPC compute-clusters through the use of HPC queuing systems. The cloud infrastructure is deployed based on user requirements that are abstracted from a web site and converted to RSpecs and stored in Catalogs for future utilization whenever similar requirements are needed.

Participants:

• Prasad Calyam
• Ronny Bazan Antequera, ​rcb553@mail.missouri.edu, University of Missouri
• Ray Leto

The GENI Experiment Engine and the Ignite Visualizer

We demonstrate the GENI Experiment Engine, one-click deployment of applications and services across the GENI infrastructure, and the use of this to deploy the Ignite Distributed Collaborative Visualization Engine, a collaborative visualization system for big data on any device, with instantaneous response to user requests

Participants:

• Rick McGeer, ​rick@mcgeer.com
• Andy Bavier
• Glenn Ricart
• Matt Hemmings
• Robert Krahn
• Dan Ingalls
• David Lary