home  |  about us  |  contact  
   

This page is part of the CSMD web archive and is not maintained.
Please visit csm.ornl.gov/newsite for the latest CSMD information.


 CSM Home
LDRD Proposal
 

Pioneering Development of High-Performance Circuit-Switched Networks

ORNL network

There are times when normal “high-speed” commercial networks simply aren’t good enough. When a scientist needs to move hundreds of terabytes of data from a remote supercomputer to his or her site, or the flow of data from an experiment is so great that it would swamp a packet-switched network such as the Internet (which is shared by everyone), then only a dedicated network will do. Unfortunately, buying a dedicated network for each scientist in the Department of Energy who would like them is neither practical nor cost effective. Fortunately, there is a better way.

The solution is circuit-switched networks, which can connect scientists to their computers with circuits that can be dedicated to any one user only as long as needed, then switched to another user. When a circuit is dedicated, it doesn’t have to use the normal network protocols that guarantee fair and equitable shared service to all users. Instead, the circuit can use greedy protocols that wring all possible performance from the network for the duration of the connection. The performance gain from using these greedy protocols can be as much as 20 times the normal network performance.

The techniques for reserving circuits, automatically setting them up and tearing them down, and connecting them to users have been developed by the UltraScience Net testbed, funded by the Department of Energy’s Office of Science through the Office of Advanced Scientific Computing Research. This testbed consists of two 10 Gigabit/second circuits that span the country, connecting Atlanta, Oak Ridge, Chicago, Seattle, and Sunnyvale. At each of these sites, switches allow connections of up to 20 Gigabits/second to users who have reserved it through a bandwidth reservation system.

The results of the testbed (which showed as much as 20 times better performance than previous connections) have been so successful that the network architecture is now being adopted by DOE’s ESnet, Internet2, and the Large Hadron Collider network. Further, it is being considered by DOD. ESnet and Internet2 currently support DOE large-scale high energy science applications. It is widely expected that very large commercial users such as banks, brokerage houses, and movie studios will also be adopting commercial versions of the same technology.

For more information, please contact:

William Wing

 

 
   CSM Projects   
   Advanced Simulation Capability for Environmental Management (ASCEM)   
   The Center for Simulation of RF Wave Interactions with Magnetohydrodynamics (SWIM)   
   Coordinated Infrastructure for Fault Tolerant Systems (CIFTS)   
   Hybrid Multi-Core Consortium   
   Integral Equation Technology   
   MADNESS (Multiresolution Adaptive Numerical Environment for Scientific Simulation)   
   NEAMS Integrated Computational Environment (NiCE)   
   Nuclear Energy Advanced Modeling and Simulation (NEAMS)   
   Reliability, Availability, and Serviceability (RAS) for Petascale High-End Computing and Beyond   
  INCITE Allocated Projects  
   Advanced Simulations of Plasma Microturbulence at the Petascale and Beyond   
   Cellulosic Ethanol: Simulation of Multicomponent Biomass System   
   Climate-Science Computational Development Team: The Climate End Station II   
   High-Fidelity Simulations for Advanced Engine Combustion Research   
   High Fidelity Tokamak Edge Simulation for Efficient Confinement of Fusion Plasma   
   Investigation of Multi-Scale Transport Physics of Fusion Experiments Using Global Gyrokinetic Turbulence Simulations   
   Magnetic Structure and Thermodynamics of Low Dimensional Magnetic Structures   
   Nuclear Structure and Nuclear Reactions   
   Performance Evaluation and Analysis Consortium End Station   
   Petascale Modeling of Chemical Catalysts and Interfaces   
   Three Dimensional Simulations for Core Collapse Supernovae   
   Ultrascale Simulation of Basin-Scale CO2 Sequestration in Deep Geologic Formations and Radionuclide Migration using PFLOTRAN   
   Uncertainty Quantification for Three-Dimensional Reactor Assembly Simulations   
   Understanding the Ultimate Battery Chemistry: Rechargeable Lithium/Air