The HOPI Project

1
The HOPI Project

• Linda Winkler
• Argonne National Laboratory
• winkler_at_mcs.anl.gov
• HOPI Design Team Co-Chair
• GNEW2004
• CERN, Geneva, Switzerland
• March 16, 2004

2
Outline

• The HOPI Project
• Problem Statement
• Goals
• HOPI Resources
• Abilene
• Internet2 10G Wave across NLR
• MANLAN Experimental Facility
• RONs
• Current Status

3
(Perceived) Problems Today

• Given current technology, Abilene is a viable
  shared packet infrastructure. By 2006 when the
  current UCAID Qwest contract expires will the
  available technology drive us toward another
  architecture?
• Some discipline specific networks have enormous
  bandwidth requirements
• High Energy Physics Community
• Square Kilometre Array Community
• Problems with packet infrastructures
• Variation in throughput, delay, jitter on shared
  packet infrastructures.
• Will we have 40 Gb/s or 100 Gb/s in near future?
  Or more likely n10G waves.
• Increasing demand for deterministic paths.
• Demand for more dynamic requirements on bandwidth
  and topology.

4
Looking to the Future

• Examine potential new architectures
• Advanced infrastructures
• Packet-based network to support basic
  functionality and application development.
• A pool of optical switched waves that support
  creation of dynamic networks for specific users
  or disciplines.
• Examining a Hybrid architecture of dynamically
  provisioned Optical lambdas and IP Packet
  Infrastructure

5
Deployment Issues with Potential New
Infrastructures

• How do you define deterministic behavior?
• How close to end nodes should optical waves be
  switched? Where is the new edge?
• How dynamic must optically switched paths be?
• What duration do we expect of dynamic paths?
• What is the interaction between the packet
  infrastructure and the dynamically created paths
  or the networks they form?
• May require new measurement and troubleshooting
  paradigm.
• What are the interconnectivity requirements to
  end systems, other research infrastructures (for
  example, international or federal networks or
  infrastructures)?

6
HOPI Project Design Team

• Rick Summerhill, Internet2 (CoChair)
• Linda Winkler, Argonne (CoChair)
• Peter ONeil, NCAR
• Bill Owens, NYSERNET
• Mark Johnson, MCNC
• Tom Lehman, ISI
• Philip Papadopoulos, UCSD
• David Richardson, PNWGP
• Chris Robb, Indiana
• Sylvain Ravot, Caltech
• Jerry Sobieski, MAX
• Steven Wallace, Indiana
• Bill Wing, ORNL
• Cees de Laat, uVa
• Rene Hatem, Canarie
• Internet2 Staff- Guy Almes, Heather Boyles, Steve
  Corbato, Chris Heermann, Christian Todorov, Matt
  Zekauskas

7
HOPI Resources

• MPLS over Abilene 10G backbone
• Internet2 10G Wave on the NLR footprint
• MANLAN Experimental Facility
• Ethernet Switch layer 2 switching
• ONS Switch layer 1 switching
• The Regional Optical Networks RONs
• Working on interface definition

8
Abilene-NLR Overlay
9
MANLAN Experimental Facility

• MANLAN Proper is an Ethernet Exchange Point for
  RE Networks
• Addition of Cisco 15454 to support projects such
  as the Global Lambda Integration Facility (GLIF)
• Current IEEAF/Tyco Circuit between Surfnet and
  Abilene will be used for international
  experimentation
• Connections
• CAnet
• Abilene
• Surfnet
• Some near term tests Lightpath consisting of a
  variety of technologies crossing multiple
  administrative domains
• There are many other experimental facilities
  StarLight (including Starlight-CERN OC192 link
  with Layer 2 and 3 services), NetherLight, etc.

10
Leading Emerging Regional Optical Initiatives

• California (CALREN)
• Colorado (FRGP/BRAN)
• Connecticut (Connecticut Education Network)
• Florida (Florida LambdaRail)
• Indiana (I-LIGHT)
• Illinois (I-WIRE)
• Maryland, D.C. northern Virginia (MAX)
• Michigan
• Minnesota
• New York New England region (NEREN)
• North Carolina (NC LambdaRail)
• Ohio (Third Frontier Network)
• Oregon
• Rhode Island (OSHEAN)
• SURA Crossroads (southeastern U.S.)
• Texas
• Utah
• Wisconsin

11
HOPI Basic Service

• Probably a 1 GE end-to-end path with reasonable
  jitter and latency properties
• Dynamically provisioned understand degree to
  which temporal setup is needed
• Understand topological extent for example,
  host-to-host? Cluster-to-cluster?
• Examine Backbone, RON, Campus hierarchy
  understand interface options
• Network operations
• Measurement activities

12
HOPI Testbed Potential Model
13
Current Status

• Design team furiously working on definitions for
• Basic services model
• HOPI node design
• Connector interface
• Control plane requirements
• Management requirements
• Target draft whitepaper for Spring (Apr)
  Internet2 Members meeting
• Planned testbed(s) implementation this year.

14
Conclusions

• Support
• Internet2 goal is to support the bandwidth needs
  of the community as networking infrastructures
  evolve in the future through shared,
  high-performance packet networks and dedicated
  capabilities as required.
• More Information
• http//abilene.internet2.edu
• http//abilene.internet2.edu/observatory
• http//www.nationallambdarail.org
• http//www.fiberco.org
• A HOPI site to be announced in the near future
• abilene_at_internet2.edu

15
(No Transcript)