Home » Previous Testbed Research

Previous Testbed Research

The following projects are no longer active.

Project Name/PI	Optimization of HEP Data Transfers on 40G NICs, Azher Mughal, Caltech
Project Summary	Determining optimal tuning, parallel streams, file system layout, and so on for 40Gbps Hosts
Expected Results	A better understanding of the perfomance tuning to saturate the 40 NIC using FDT.

Project Name/PI	NERSC 100Gbps DTN testing and tuning, Damian Hazen, NERSC
Project Summary	Optimizing NERSC DTN Nodes to fill the 100G Network from GPFS
Expected Results	A better understanding of the perfomance tuning for NERSC DTNs.

Project Name/PI	End-system affinities and the performance of high-speed flows, Dipak Ghosal (UC Davis)
Project Summary	Characterizing the impact of end-system affinities on the end-to-end performance of high-speed flows.
Expected Results	A better understanding of the perfomance impact of core selection, and more.

Project Name/PI	Analysis of UDT/UDR at 40G, Bob Grossman (University of Chicago)
Project Summary	Perform performance testing of the UDR data transfer tool over high-speed networks.
Expected Results	A faster UDR tool, and better understanding of the tuning required for optimal perfomance.

Project Name/PI	Web10G TCP Extended Statistics Performance at 40Gbps, Chris Rapier (PSC)
Project Summary	Perform performance and stability tests on the web10G Linux kernel.
Expected Results	A stable web10G kernel that can be used to instrument TCP flows at rates up to 40Gbps.

Project Name/PI	Advanced cyber-security intrusion detection at 100Gbps speeds / Alan Commike, Reservoir Labs
Project Summary	Validate and tune algorithms and implementations associated with scaling a network intrusion detection system to 100G networks.
Expected Results	Ability to run a full featured network intrusion detection system at 100G speeds within a reasonable size, weight, and power budget. Knowledge gained will enable further development to push well beyond 100G traffic rates.

Project Name/PI	Large Data Access and Transport Proposal, Linden Mercer (NRL) and Soumik Sinharoy (Orange Silicon Valley)
Project Summary	The proposed work will explore extension of previous accomplishments by NRL, OSV in delivering high performance data access and transport using RDMA transport, particularly InfiniBand and clustered file systems, particularly Lustre.
Expected Results	Validation of the ability to do single host, disk-to-disk, 100Gbs using RDMA, and explore the performance limits and tuning for the Lustre file system.

Project Name/PI	Usability Investigations for High Energy Physics Analysis: Ruth Pordes, FNAL (renewal)
Project Summary	This project will study the issues related to end-to-end integration and use of 100Gigabit networks for the Grid middleware applications and services used by the High Energy Physics (HEP) communities.
Expected Results	We will use the testbed for data access tests of existing HEP Physics middleware and applications (e.g. Using gridFTP, dCache, NFS, Globus Online, SQUID, Frontier data caching, and US CMS XROOTD based data caching when available).

Project Name/PI	Climate100: Enabling 100Gbps for Climate Community / Alex Sim (LBNL) (renewal)
Project Summary	To integrate massive climate change simulation datasets and the 100Gbps network transport and management technologies, we'll study the effectiveness and readiness of the data transfer protocols over 100Gbps in the context of climate data characteristics, and prepare climate community to transition to the 100Gbps network for production.
Expected Results	To improve the understanding and use of the new technologies in network, storage and file system, and help the climate community transition to the 100Gbps network for production and research.

Project Name/PI	HNTES: Malathi Veeraraghavan, University of Virginia (renewal)
Project Summary	This project will design, implement, and experiment with a system called Hybrid Network Traffic Engineering Software (HNTES). The purpose of HNTES is to leverage both an IP datagram network and a high-speed optical dynamic circuit network to best serve users' data communication needs. It will enable scientific teams to leverage the value of high-speed optical dynamic circuit networks without requiring modifications to their current applications that were originally programmed for IP datagram networks.
Expected Results	Experiments on the testbed will be conducted to determine whether flows can be redirected on-the-fly to newly established optical circuits without impacting TCP behavior, and user-perceived performance.

Project Name/PI	Loss Patterns and Loss Recovery at 100Gbps: Lisong Xu, University of Nebraska-Lincoln
Project Summary	It has been observed that even a very small amount of packet loss can have a huge impact on performance over high-latecy paths in ESnet. The poor performance of high-speed networks with even small packet loss rates motivates us to study loss patterns and loss recovery at 100Gbps.
Expected Results	We attempt to answer the following two questions in this project. 1) what are packet loss patterns in 100Gbps networks? 2) What is the loss recovery efficiency in 100Gbps networks?

Project Name/PI	End-to-end Data-flow Parallelism for Throughput Optimization on 100 Gbps Networks: Tevfik Kosar, University at Buffalo
Project Summary	To analyze different factors that affect the end-to-end data transfer throughput on a 100 Gbps network, such as number of parallel streams, buffer size, CPU and I/O speeds at the end systems (source and destination). The experiments will be conducted using both TCP and UDT.
Expected Results	To show the effects of CPU-, disk-, and network-level parallelism in removing the end-to-end performance bottlenecks one-by-one and come up with a generic model for optimizing the end-to-end data transfer throughput.

Project Name/PI	Scaling Network Protocols and Applications to 100Gbps and beyond, Bill Fink, NASA Goddard
Project Summary	Research and prototype the synthesis of end system hardware (motherboards, processors, memory, system buses, NICs, raid controllers, spinning/solid-state disks), operating system kernels and software drivers, network protocols, file transfer applications, and network components (optical equipment, routers, switches, and links), that is required to achieve 100 Gbps end-to-end disk-to-disk network data transfers at full line rate on an economical high-end server system.
Expected Results	Determine how to effectively integrate the necessary parallelism across RAID disk arrays, multi-core CPU architectures, parallel TCP (or UDP) data streams, multi-queue 100G NICs, and 10/40/100G network infrastructure, to fulfill the goal of 100 Gbps end-to-end disk-to-disk network data transfers. Demonstrate an implemented capability that approaches the desired goal, by running a network data transfer application between a single pair of economical high-end server systems across the real 100G WAN environment that is provided by the ANI testbed. Evaluate the scalability of the hardware, software, and methodology for networks beyond 100 Gbps.

Project Name/PI	Evaluating RAPID TCP: Jasleen Kaur, University of North Carolina
Project Summary	Experimentally evaluate the software and hardware prototypes for RAPID TCP that are being developed in our group.
Expected Results	The goal of our evaluations will be to: (i) measure the degree of buffering-related noise encountered by fine-scale probing on ultra-high speed network paths, (ii) evaluate the efficacy of the paradigm in dealing with the noise, and (iii) evaluate the efficacy of the prototypes in controlling inter-packet gaps at ultra-high network speeds.

Project Name/PI	Scaling GridFTP to 100Gbps, Raj Kettimuthu, ANL
Project Summary	The goal is to evaluate the performance of GridFTP on ANI using multiple transport protocols including TCP and UDT with varying degrees of parallelism for datasets ranging from lots of small files to big files.
Expected Results	Identify bottlenecks and potential areas of improvement in GridFTP.

Project Name/PI	Experimentation with Resource Provisioning for Wide Area Hadoop-Based Scientific Applications, Dr. Anirban Mandal and Dr. Yufeng Xin, RENCI
Project Summary	We propose to use the ANI application/middleware testbed for experimentation with provisioning and embedding novel resource topologies appropriate for next generation, data-intensive science applications enabled by 100Gbps networks. Specifically, we plan to experiment with Hadoop based applications, which can potentially be deployed across multiple sites connected by high performance networks. We plan to evaluate the performance and scalability of Hadoop applications when deployed on the ANI 100 Gbps application/middleware testbed.
Expected Results	We plan to present comparisons of different provisioning configurations for Hadoop deployments by quantifying the performance overhead (virtualization, network connection capacities and delay) of multi-site deployments using the ANI testbed, and determining the sweet-spots for when to use more sites. Since performance of HDFS is sensitive to underlying network characteristics, we would present a comparison of single vs. cross site HDFS performance using Hadoop benchmarks.

Project Name/PI	Viability of the Intelligent Control Plane Platform based on OpenFlow enhanced with MPLS/GMPLS capabilities; Ben Yoo, UCDavis and Vinod Mishra, DISA
Project Summary	The proposed project will design and conduct testbed experiments to 1) evaluate the aggregation, provisioning, monitoring and protection/restoration of DISN traffic pattern using OpenFlow based automotive and adaptive control plane, 2) compare flow based control with MPLS/GMPLS based control to identify the differences and similarities between them, and analyze the interoperability of OpenFlow with MPLS/GMPLS, 3) design and analyze better protocols for enhanced TCP/UDP variants to support GridFTP massive data transfer.
Expected Results	We will reserve the resources on the Phase-3 Cross country 100 Gbps Testbed control plane and middleware testbed architectures to set up a 100Gbps DWDM transport platform. Different physical topologies will be set up in the testbed using the resources such as NEC OpenFlow switches, application hosts, and monitoring hosts. Experiments such as aggregation, provisioning, and monitoring of the DISN traffic mix using OpenFlow, as well as benchmark performance of GridFTP file transfers will be carried out on the testbed.

Project Name/PI	Evaluation and Performance Characterization of Saratoga, a Reliable High-Speed Transport Protocol: William Ivancic, NASA Glenn Research Center
Project Summary	Evaluation and Performance Characterization of Saratoga, a UDP based protocol, and high speed.
Expected Results	Determine the operational performance of various implementations and identify code bottle necks and potential areas that could improve performance.

Project Name/PI	VNOD: Virtual Network On Demand / Dimitrios Katramatos, BNL
Project Summary	VNOD has as main goal to provide the capability to establish guaranteed-performance, high-bandwidth virtual networks. Such networks will enable the federation of storage and computing resources on an on-demand basis.
Expected Results	We will use the testbed to Investigate techniques and algorithms for negotiating and coordinating reservations of network resources between multiple systems involved (OSCARS, ARCHSTONE, ESCPS, TeraPaths) in the configuration of a virtual network. Also, to study the efficiency and efficacy of QoS in different virtual topologies and at speeds beyond 10Gb/s.

Project Name/PI	Deploying and Validating RDMA Transports and Protocols at 40 Gigabit per second: Paul Grun, System Fabric Works
Project Summary	To deploy and validate the operation and performance of InfiniBand Transports and RDMA protocols on a WAN connection over a 240 km at 40Gigabits per second.
Expected Results	Determine if there any issues running RDMA at 40Gbps over a long distances

Project Name/PI	End-2-End Network Virtualization using OSCARS and OpenFlow based NEC ProgrammableFlow: Samrat Ganguly, NEC
Project Summary	Proposed research is to create and demonstrate for the first time a programmable end-to-end wide-area network virtualization infrastructure.
Expected Results	The goal of our evaluations will be explore various ways to integrate OSCARS with OpenFlow and determine what works best.

Project Name/PI	Disk/Network Bottleneck Detection at 10/100G: Dan Gunter, LBNL
Project Summary	We have developed a framework for streaming analysis of user transfers that can efficiently determine, at gigabit rates, whether the bottleneck is at the source disk, network, or destination disk. We plan to validate these algorithms at the higher data rates available on the ANI testbed.
Expected Results	Validation and improvement of bottleneck detection algorithm.

Projects added Febuary 2011

Project Name/PI	Advance Scheduling of Multi-Domain Dynamic Circuits: Byrav Ramamurthy, University of Nebraska-Lincoln
Project Summary	This project will investigate multi-domain dynamic circuit creation using ANI testbed. We will study the issues related to large data transfers over multi-domain circuits.
Expected Results	The expected result of this project will be additional features in the existing control plane architecture of OSCARS based on our experiments involving intra-domain and multi-domain dynamic circuit creation over the ANI testbed. The issues related to transfer of large amounts of data over dynamic circuits established across multiple domains will be studied in detail.

Project Name/PI	Usability Investigations for High Energy Physics Analysis: Ruth Pordes, FNAL
Project Summary	This project will study the issues related to end-to-end integration and use of 100Gigabit networks for the event simulation and analysis applications of physics experiments.
Expected Results	We will use the testbed for data access tests of existing HEP Physics middleware and application components (e.g. Using LUSTRE, Frontier data caching, and US CMS XROOTD based data caching when available). We will write reports based on the results to give recommendations to the system administrators, application and middleware developers on changes that would make production use of ANI more effective, including data storage, caching and wide area access.

Project Name/PI	Securing Network Services using DASH: Ben Smith, Angel Secure Networks
Project Summary	ANGEL Secure Networks, Inc.'s DASH product uses a network of software agents to defend critical software systems from insider and outsider attack and tampering. DASH was originally developed for DoD applications, running on networks of processors in cluster computers and embedded systems. We have a DOE Phase II SBIR to apply our technology to high-performance networks. We plan to use the ANI testbed to acquire a better understanding of the type of networks used for DOE science and to demonstrate how our system can help to protect them.
Expected Results	We will demonstrate a network of our software agents (ANGELs) protecting routers on the ANI testbed under a variety of attack scenarios.

Project Name/PI	Measuring Energy Efficiency In Networks: Thierry E. Klein, Bell Labs / Alcatel-Lucent
Project Summary	The main purpose of this project is to gain an understanding of the power-rate profile and the energy efficiency in real-world routers, switches and networking equipment and to explore opportunities for improving energy efficiency through dynamic management and control (including rate adaptation and sleep modes).
Expected Results	The project will determine the power-rate profiles of the available equipment in the testbed, and in particular, determine the power-rate profiles of the Juniper MX80 layer 3 routers and the Infinera layer 2 switches. In addition, we will conduct experiments turning the equipment on and off to understand the behavior, the state transition times and power consumption during such operations. Finally we will investigate the implementation of sleep modes and rate adaptation and quantify the obtained energy benefits. The feasibility of the latter experiment depends on the hardware capabilities and our ability to modify their operations.

Project Name/PI	Testing high speed protocol PERT over a real 10Gbps network: Narasimha Reddy, Texas A&M University
Project Summary	This project will test PERT TCP on a real 10Gbps network over long distances.
Expected Results	A comparison of PERT to TCP-SACK and other high-speed protocols over single flow performance and live video delivery performance.

Project Name/PI	Scalable Optical Networking with OpenFlow : S. J. Ben Yoo, UC Davis
Project Summary	This project will design and conduct testbed experiments of OpenFlow based future ESnet. In anticipation of increasing traffic demands and required needs to support various levels of end-to-end quality of service, DOE is in need of intelligent and agile network infrastructures with 1) scalable optical networking, 2) hybrid packet/circuit-switched networking, and 3) multi-layer multi-domain network measurement and monitoring.
Expected Results	This project will investigate, develop, test, and help standardize OpenFlow towards supporting scalable and dynamic optical networking. Support of packet and circuit- based services, dynamic flow aggregation/ deaggregation, cross- layer/cross-domain network monitoring, control and management are expected from this project.

Projects added June 2010

Project Name/PI	Climate 100: Alex Sim, LBNL
Project Summary	The Climate100 project integrates massive climate datasets, emerging 100Gbps networks, and state-of-the-art data transport and management technologies. The goal of this project is to improve the understanding and use of network technologies and transition the climate community to a 100 Gbps network for production and research.
Expected Results	The testbed will be used to test the direct memory access over the network and new data transfers/management algorithms including the use of the 100G transfer protocol.

Project Name/PI	ARCHSONE: Tom Lehman, ISI
Project Summary	The goals of the ARCHSTONE project are to develop technologies which enable resource computation and provisioning across next-generation multilayer network architectures. The overall goal is to dynamically create “slices” of resources across multiple network layers in a vertically integrated manner, so as to generate virtual network topologies, circuits, and/or partitions to support various services in a flexible fashion. A key focus will be development of an enabling technology known as a “Multi-layer/Multi-dimensional Resource Computation Element”(MX-RCE). This entity will effectively serve as a highly-advanced path computation element which extends the concept of simple path computation to multilayer, multidimensional topologies.
Expected Results	Based on the experiment results, we plan to write a report on the observed performance and any bottleneck scenarios experienced on 100Gbps network paths in the ANI testbed. We also expect to review capabilities and gain familiarity with the Openflow and 100Gbps network device technologies.

Project Name/PI	End-System Network Interface Controller for 100 Gb/s Wide Area Networks: Jesse Wen, Acadia
Project Summary	This project will develop network interface controller (NIC) hardware and device-driver/protocol-specific software for host and gateway systems operating at 40 and 100 Gb/s.
Expected Results	The testbed will be used to investigate issues that do not arise in initial back-to-back testing. Such issues include interoperability with core-network gear and the effect of long-haul physical impairments.

Project Name/PI	100G FTP: An Ultra-High Speed Data Transfer Service : Dantong Yu, BNL
Project Summary	This project will design and develop an ultra high speed end-to-end file transfer protocol and tool to move science data at a speed of 100Gbps across the national scale 100Gbps data network interconnecting research centers.
Expected Results	The testbed will be used to verify that this tool scales to 100Gbps on a single wavelength or multiple modulated wavelengths.

Project Name/PI	Advanced Network and Distributed Storage Laboratory (ANDSL): Haifeng Pi, UCSD
Project Summary	To enhance Virtual Data Toolkit (VDT) data management tools to effectively utilize 100G networks.
Expected Results	The testbed will provide interconnect between “Magellan project” resources at ANL and NERSC.

Project Name/PI	FlowBench: Prasad Calyam, Ohio Supercomputer Center
Project Summary	We will set up different physical topologies in the ANI testbed using the testbed resources such as NEC Openflow switches, App Hosts, and Monitoring hosts. On these topologies, we plan to experiment with Openflow and benchmark performance of GridFTP file transfers with enhanced TCP/UDP variants.
Expected Results	Based on the experiment results, we plan to write a report on the observed performance and any bottleneck scenarios experienced on the network paths in the ANI testbed. We also expect to review capabilities and gain familiarity with the Openflow and 100Gbps network device technologies.