Completed Research Projects


Boosting Inter-Domain Scheduled Dynamic Circuit Services (SDCS)

Funding agency: CNS, Division Of Computer and Network Systems
PI: Malathi Veeraraghavan. Award: 1116081 $250,000. August 1, 2011 – July 31, 2016.

A variety of applications in the nascent field of cloud computing and in other fields requires stable, low-delay network connectivity together with capabilities of co-scheduling network and server resources. As a result, several network providers have recently started to deploy scheduled dynamic circuit services (SDCS). Under such services, a network provider allows its customers to place scheduling requests for a fixed-rate circuit lasting for a fixed duration, for either earliest possible usage or scheduled usage at a desired point of time in the future.

While carrying significant potential impact, SDCS currently provide little support for dynamic routing and scheduling of circuits across different administrative domains. This lack of adequate support for inter-domain routing represents a major hurdle to the full-scale deployment of SDCS. The main objective of this project is to tackle this challenge by devising distributed solutions supporting scheduled dynamic circuit services across domains. The project consists of two inter-related thrusts: (i) theory, which focuses on the problem of devising distributed routing algorithms for SDCS that provably converge and provide performance guarantees on user-performance metrics; and (ii) protocol design and implementation, which focuses on the design, development, and real network prototyping of a new inter-domain routing protocol for SDCS, referred to as the scheduled circuit routing protocol (SCRP).

Broader Impact: Successful completion of this project promises to impact a wide array of high-throughput and real-time applications belonging to medical, scientific and commercial domains, and facilitate the creation of a new ecosystem for communication networks. The project is further striving for high impact through collaborations with international academic and industrial partners, and network operators. Students at all levels will be engaged in real network experiments and in the development of innovative applications enabled by SDCS, such as haptic-based communication.

 

An integrated study of datacenter networking and 100 GigE wide-area networking in support of distributed scientific computing

Funding agency: ACI, Division of Advanced Cyberinfrastructure
PI: Malathi Veeraraghavan. Award: 1127340 $498,378. September 15, 2011 – August 31, 2016.

As supercomputing speeds increase to peta- and exaflops, scientists are increasing their scale and range of simulations, which are resulting in ever growing datasets that need to be moved to local computers at the scientists’ own laboratories. The first goal of this project is to identify bottlenecks that result in poor and/or inconsistent end-to-end application-level throughput using data collection and analysis by working in conjunction with scientists in the Community Earth System Model (CESM) project. With knowledge of the weakest components in the end-to-end chain, we plan to experiment in a controlled environment using a testbed that consists of a high-end cluster at NERSC, which is capable of sourcing/sinking data to disks at close to 100Gbps speeds, and other high-performance computing systems connected via the DOE 100Gbps Advanced Networking Initiative (ANI) prototype network. Multiple datacenter networking technologies such as Remote Direct Memory Access (RDMA) over Converged Ethernet (RoCE) and Internet Wide-Area RDMA Protocol (iWARP) will be combined with high- speed (100Gb/s) wide-area networking solutions, such as dedicated virtual circuits and IP-routed paths, respectively, for a comparative performance study of file transfers and wide-area MPI I/O. A new software module of the Extended-Sockets API (ES-API), which offers RDMA features such as zero-copy operations, will be prototyped and integrated into file transfer applications. Finally, trials will be organized to transfer the best identified solutions to CESM and other scientists. The intellectual merit of the proposed project consists of: i) a systematic scientific approach to determine the reasons for poor end-to-end application-level performance experienced by CESM scientists, ii) development of integrated datacenter and wide-area networking solutions to address the identified problems, and iii) the enabling of these solutions to be utilized by CESM and other science projects. The broader impacts of the proposed activities consist of i) the creation of a course module on datacenter networking, and the involvement of undergraduate students in this research at all three institutions, ii) diversity and outreach programs, and iii) the active promotion of the developed solutions to the CESM project and other scientists.

 

Towards increasing the usage of new high-speed network services by the scientific community

Funding agency: NSF Division of Advanced Cyberinfrastructure
PI: Malathi Veeraraghavan. Award: 1038058 Eager $299,916. September 1, 2010 – August 31, 2013.

The proposal is high-risk, high payoff because it is addressing a fundamental networking problem which has used “overprovisioning” IP as the primary way to address congestion issues. While this approach has been (and continues to be) used with considerable success, it is only part of the solution; however, because of the embedded base of using this approach, changes to the IP stack or protocol are strongly discouraged. This approach is particularly problematic for data applications that are characterized by long data flows and streams.

This proposal takes a fresh look at going beyond ubiquitous IP datagram services and considers circuit/virtual circuit networking alternatives. The effort will use the Unidata LDM application to demonstrate and implement a modified version of the proposed network to show improvements in performance as well as a reduction in costs. The proposal will undertake an in-depth study of LDM and its associated Internet Data Distribution project to test the feasibility of off-loading long duration data flows from the general IP network to virtual-circuit options.

 

End-to-end Provisioned Optical Network Tested for Large-Scale eScience Applications

Funding agency: NSF Division of Advanced Cyberinfrastructure
PI: Malathi Veeraraghavan. Award: 0335190 $2,113,000. January 1, 2004 – June 30, 2009.

Project Website

This proposal is a comprehensive effort to develop the infrastructure and networking technologies to support a broad class of eScience projects and specifically the Terascale Supernova Initiative. The proposed work will build a high-performance, experimental, optical network infrastructure that supports: (a) on-demand provisioning of end-to-end high-speed circuits, (b) stable transport to sustain control and streaming operations, and (c) middleware and application software to support data transfers, visualization, steering and control.

Intellectual Merit: Through vertical integration of these components, the researchers plan to demonstrate how a large-scale eScience application, such as TSI, can fully exploit the benefits of optical networks. The team consists of optical networking researchers, transport-layer and middleware researchers, and radio-astronomy scientists. Successful completion of this project will allow the Awardee to determine the benefits and costs of creating a backbone high-speed shared circuit-switched network along the same lines as the two packet-switched Internet2 backbone networks, Abilene and vBNS.

Broader impact: Due to the broad spectrum of target eScience applications, the project should have a broad societal impact. Given the diverse team of researchers on this project, results should be disseminated in various research communities rapidly and increase the level of interest in this approach. This project significantly leverages various resources at the universities and ORNL. Students from these universities will be provided access to computational and network resources at ORNL, in terms of supercomputers and high-bandwidth links. Also, this work will be used to provide projects to students from DOE educational programs, which support undergraduate and graduate students from universities across the country. This project will specifically include summer students from ORNL’s Research Alliance for Minorities program.

 

Fast File Transfers Across Optical Circuit-Switched Networks

Funding agency: NSF Division of Advanced Cyberinfrastructure
PI: Malathi Veeraraghavan. Award: 0312376 $324,999. September 15, 2003 – August 31, 2007.

Project Website

The PIs propose a network solution called Dynamically Reconfigurable Ethernet/Ethernet-over-SONET (DREEoS) to enable end-to-end connectivity through high-speed optical circuits. The primary application that is considered in this project is large data transfers between businesses (B2B), and between businesses and consumers (B2C). DREEoS is proposed as an add-on service to the primary Internet access service. This can be achieved by equipping end hosts with second (high-speed) Ethernet cards and connecting these cards to ports on an enterprise Multi-Service Provisioning Platform (MSPP). Wide-area SONET circuits are established dynamically between enterprise MSPPs and Ethernet signals from the end hosts to the enterprise MSPPs are mapped to these wide-area circuits. An end host with access to DREEoS service has to determine whether it is worthwhile attempting a DREEoS circuit or not. Through analysis, it showed that an end host should first attempt setting up a DREEoS circuit if (i) file sizes are large (ii) file sizes are small but round-trip times (RTT) are large and (iii) file sizes and RTT are small, then for files larger than some crossover file size. The crossover file size and crossover RTT depend upon the probability of packet loss on the TCP/IP path, link rates, RTT, and call blocking probability on the optical circuit-switched path. The availability of the fallback TCP/IP path allows DREEoS service to be introduced gradually into optical networks. At low loads, the network can be operated at high call blocking probabilities to achieve high utilization. As loads increase, the network can be engineered to retain high utilization while simultaneously offering low call blocking probabilities. In this project, the PIs plan to implement various software modules, such as application software to trigger circuit setup, Optical Connectivity Service (OCS) to allow an end host initiating a transfer to determine whether its correspondent host can even be reached via a DREEoS circuit, transport protocol software to allow for high-speed transfers, and integrate these modules to demonstrate high-speed file transfers across the university campuses of the PIs.

This work represents a significant advance in our current understanding of the potential of circuit-switched networks. By using hardware-accelerated signaling engines the network can support calls with short holding times. This means applications can exploit the delay benefits of circuit-switched networks, critical for interactive applications, without compromising utilization. This work will reopen our thinking on circuit-switched networks enabling new research. For example, the PIs are considering replacing the call blocking mode used in this proposal with a call queueing/scheduling mode. This will remove the impact of propagation delay by allowing for staggered configuration of switches, a feature that is important for very small call holding times. This work has a clear application in Science and Engineering, where distant collaborations require massive file transfers in short durations. Our educational and diversity goals will be met through the involvement of students, including minorities and women, from both engineering and management schools. The multidisciplinary nature of this project will help prepare our students better for future industrial jobs. The expected major contributions of the proposed research include (i) the dissemination of our software modules, which will be useful to other research teams working on similar optical network concepts, (ii) demonstrations of the viability of this service model for commercial deployment through our economic and market analysis, and (iii) research papers enhancing the role of optical circuit- switched networks.

Resource Optimization in Hybrid Core Networks with 100G Links

Funding agency: US DOE
PI: Malathi Veeraraghavan. $524,806. 2009-2013.

Project Website

 

Terabit-scale hybrid networking

Funding agency: US DOE
PI: Malathi Veeraraghavan. $497,442. 2012-2015.

Project Website

 

Enabling Supernova Computations by Integrated Transport & Provisioning Methods Optimized for Dedicated Channels

Funding agency: US DOE
PI: Malathi Veeraraghavan. $240,000. 2004-2006. UVA (sole PI) portion: $240,000.

 

Advance Freeway Merger Assistance: Harnessing the Potential IntelliDrive

Funding agency: US DOE
My role: co-PI. $500,000. Sep. 2009-Aug. 2011.

Project Website

 

NSF: Hardware acceleration of signaling protocols

Support: NSF ANIR. Project number 0087487.

Project Website

  • UVA Teaching Resource Center Fall 2012 Challenge for Newly Hybrid TechnologyEnhanced Courses, Computer Networks Flipped-Classroom Offering, $10,000, my role: PI, 2012-2013
  • Internet2/Cisco, Equipment grant of two Cisco 12008 high-end IP routers, Dec. 2002; rough estimate: $100,000 (UVA; sole PI).
  • NSF, Towards enabling a 2-3 orders of magnitude improvement in call handling capacities of switches, $450,000, plus $60,000 in cost-share support from Polytechnic University, 2001-2006, my role: PI; my portion: $405,000 of which $165,285 was moved to UVa (Sole PI) in 2004.
  • Village Networks, A study of different optical metropolitan-area network architectures, $116,166, 2001-2002, my role: PI; Sole PI.
  • Verizon, Automation of SS7 testing, $262,697, Sept. 2000 – Aug. 2001, my role: PI. Sole PI.
  • Verizon, SS7 testing methodology development, $170,000, Sept. 99-Aug. 00, my role: PI. Sole PI. (Project Website)
  • NYSTAR (New York Office of Science, Technology and Academic Research) through the Center of Advanced Technology in Telecommunications (CATT), approx. $548,863. Sole PI.
  • NIST, Mobile Information Infrastructure, 1994-1997, $12 Million, Bell Labs, my role: co-PI. Large team.