Upload
yolanda-sullivan
View
24
Download
0
Tags:
Embed Size (px)
DESCRIPTION
Internet2. Intel Partnership Planning Meeting November 19,2001. What Is Internet2?. A project of the university community working with our corporate colleagues and government to close the gap between the potential and reality of the Internet. - PowerPoint PPT Presentation
Citation preview
2
What Is Internet2?
A project of the university community working with our corporate colleagues and government to close the gap between the potential and reality of the Internet
4
Internet2 Mission
Develop and deploy advanced network applications and technologies, accelerating the creation of tomorrow’s Internet.
5
Internet2 Goals
Enable a new generation of applications
Re-create leading edge R&E network capability
Transfer technology and experience to the global production Internet
6
Today’s Internet Doesn’t
Provide reliable end-to-end performance
Encourage cooperation on new capabilities
Allow testing of new technologies
Support development of revolutionary applications
7
Why University Leadership?
The Internet came from the academic community
• Stanford -- the Internet protocols• NSFNet -- the scaled-up Internet• CERN -- the WWW protocols• University of Illinois -- the Web browser
Universities’ research and education mission require an advanced Internet and have demonstrated they can develop it
8
Technology Transfer Conduits
Collaborating on advanced applications
Deploying pre-commercial infrastructure and protocols
Establishing expertise and human capital
Large-scale proof of concept
9
Research andDevelopment
Commercialization
Partnerships
Privatization
Internet Development Spiral
Today’s Internet
Internet2
NSFNetARPANet
NYSERNet
SURANet
MichNet
ANS/Core
PSI
UUNet
InternetMCI
AOL
GigaBitTestbeds
MBone
NGIIntelligentNetworks
Source: Ivan Moura Campos
10
Internet2 Activities and Focus Areas
Advanced Network Infrastructure
Middleware
Engineering
Advanced Applications
End-to-End Performance
Advanced Network Management
Partnerships
12
Internet2 Network of the Future
Current state of Abilene
Evolution of optical networking
Next phase of Abilene
13
Abilene background & milestones
Abilene is a UCAID project in partnership with• Qwest Communications• Nortel Networks• Cisco Systems• Indiana University• ITECs in North Carolina and Ohio
Timeline• Apr 1998: Project announced at White House• Jan 1999: Production status for network• Oct 1999: IP version of HDTV (215 Mbps) over Abilene• Apr 2001: First state education network added• Jun 2001: Participation reaches all 50 states & D.C.• Nov 2001: Raw HDTV/IP (1.5 Gbps) over Abilene
14
Abilene focus
Enabling innovative applications and services not possible over the commercial Internet
Advanced service efforts• Multicast• IPv6• QoS• Measurement• Security
– DDoS detection efforts (Arbor Networks & Asta Networks)
15
Abilene status – November, 2001
IP-over-SONET (OC-48c) backbone
54 direct connections• 3 OC-48c (2.5 Gbps) connections • 22 will connect via at least OC-12c (622 Mbps) by year end
200+ primary participants • All 50 states, District of Columbia, & now Puerto Rico• 15 regional GigaPoPs support ~70% of participants• 37 sponsored participants
15 state education networks (SEGPs)• Collaboration of sponsoring member universities and
Abilene connectors
18
International peering
Transoceanic R&E bandwidths growing!
Key international exchange points facilitated by Internet2 membership and the U.S. scientific community
• STARTAP STAR LIGHT – Chicago • Pacific Wave – Seattle • AMPATH – Miami• New York City – EP under development
• CUDI - CENIC and Univ. of Texas at El Paso
International transit service
19
Measurement and DDoS
Traffic characterization (Ohio ITEC)• Network utilization by SEGPs and Abilene ITN• Abilene Scavenger Service policing• GigaPoP pair hotspot identification
Passive measurement • Planned for Indianapolis router backbone links• Collaboration with SDSC
Distributed Denial of Service detection• Strong IU Global NOC interest• Asta Networks (UCSD/U of Washington roots)• Arbor Networks (U of Michigan/Merit roots)
Data privacy and anonymity policy
20
Network of the Future:Context for the next backbone
Computational science as an emerging interdisciplinary field
• Bandwidth and distributed sensing capability as the next critical parameters
– Complement CPU, memory & storage
• Increasingly distributed data collection and storage• NSF Distributed Terascale Facility solicitation
Emergence of optical technologies• Dense Wave Division Multiplexing (DWDM)• Important distinction: optical transport vs. switching
Much new transcontinental conduit and fiber in place; a lot of business plans abandoned…
• Glut of fiber & conduit – but not bandwidth
21
Current state of optical networking
Dense Wave Division Multiplexing (DWDM) • Current systems can support >160 10-Gbps ’s (1.6 Tbps!)• Optical growth can overwhelm Moore’s Law (routers)
Costs scale dramatically with distance
Three possible scenarios for the future• Enhanced IP transport (higher BW and circuit multiplicity)• Fine-grained traffic engineering
– p2p links between campuses, HPC centers, & Gigapops
• Physical manifestation of switched circuits (a la ATM SVCs)
Evolution of optical switching will be critical
Leading international efforts in R&E exploration• The Netherlands, Canada, STAR LIGHT (Chicago)
22
National optical networking options
1 - Incremental wavelengths• Provision 10-Gbps ’s from provider(s) in the same way
that SONET circuits are done for Abilene now• Exploit smaller incremental cost of additional ’s
2 - Dim Fiber• Acquisition of fiber IRU and subsequent O&M
agreement for inter-PoP services (amps, regenerators, DWDMs?)
• National footprint of 1-2 fiber pairs – IRU would cost $10-20M
• Most likely awaits the availability of lower-cost optical transmission equipment
23
Future of Abilene
Original UCAID/Qwest MoU amended on October 1, 2001
Extension of Qwest’s original commitment to Abilene for another 5 years – 10/01/2006
• Originally expired March, 2003
Upgrade of Abilene backbone to optical transport capability - ’s
• x4 increase in the core backbone bandwidth– OC-48c SONET (2.5 Gbps) to 10-Gbps DWDM
• Capability for flexible provisioning of ’s to support future point-to-point experimentation & other projects
24
Key aspects of the next backbone
IPv6• Running natively concurrently with IPv4• Replicate multicast deployment strategy• Motivations
– Resolving IPv4 address exhaustion issues– Preservation of the original End-to-End Architecture– International collaboration– Router and host capabilities
• Close collaboration with Internet2 IPv6 Working Group
Network resiliency• MPLS/TE fast reroute or IP-based IGP fast convergence
Opportunity for new measurement capabilities• Support of End-to-End Performance Initiative
25
Next generation network deployment
October, 2001: Detailed technical design starts
February, 2002: PoP upgrades start
deployment in three phases• April, 2002 – Phase 1• October, 2002 – Phase 2• April, 2003 – Phase 3
October 2003 - Completion of 10-Gbps upgrade
26
Network design overview
Overall next generation topology is expected to be very similar to current design
• Previous iterations to router locations –Washington DC, Chicago, Sunnyvale, Houston
• Some differences expected due to Qwest DWDM deployment
• Expect same number of backbone routers
27
Optical fanout
Next generation architecture: Regional & state based optical networking projects are critical
• Three-level hierarchy: backbone, GigaPoPs, campuses• CENIC ONI, I-WIRE, SURA Crossroads, Indiana, Ohio• Pacific/Northwest Gigapop and PREN are relevant players
in the Northwest
Collaboration with the Quilt • Regional Optical Networking project
Carrier DWDM access is now not nearly as widespread as with SONET circa 1998
28
The Quilt
A UCAID project support regional advanced networking initiatives
• 15 charter GigaPoPs• EDUCAUSE and SURA• Quilt GigaPoPs support over 70% of Abilene participants
Initial projects• Commodity Internet Services• Regional Optical Networking• Measurement
Led by Wendy Huntoon (Pittsburgh SC)
29
Conclusions
• Abilene future• UCAID’s partnership with Qwest extended through 2006• Backbone to be upgraded to 10-Gbps in three phases
starting spring 2002• Capability for flexible provisioning in support of future
experimentation in optical networking• Overall approach to the new technical design and business
plan is for an incremental, non-disruptive transition
•Follow-on network most likely will be developed around national dark fiber facility and will utilize next generation optical transport technology
33
Internet2 Multicast
Kevin Almeroth, Univ California Santa Barbara, chair
Increasingly pervasive high-quality deployment of native IP multicast throughout the Internet2 infrastructure
Keeping an eye on SSM Implications of SSM on scalability, manageability Adapting applications to make use of SSM
Clarifying the application story Internet2's multicast infrastructure is a valuable sand
box in which to test the value of new multicast applications
35
Internet2 IPv6
Dale Finkelson, Univ Nebraska, Michael Lambert, PSC, co-chairs
Build the Internet2 IPv6 infrastructure Currently, based on v4-over-v6 tunnels Planned as first-class service with the coming 10
Gb/s upgrade of Abilene
Educate campus network engineers to support IPv6
Explore the Motivation for IPv6 within the Internet2 community
Make IPv6 'real' within the university community (and to our students)
37
Internet2 QoS
Ben Teitelbaum, Internet2 staff, chair
QBone Premium Service Hard priority service for selected streams Very hard due to need for policing/administration
Scavenger Service Voluntary less-than-best-effort service Enables unconscionable bulk data transfers without
threatening performance of best-efforts traffic
Other 'non-elevated' services E.g., delay- vs loss-sensitive best effort service Interoperability without policing / administration
38
Internet2 Measurements
Matt Zekauskas, Internet2 Staff, chair
Define architecture: Usage Active Measurements of Performance Passive Measurements
Uniform Access to Results
Contributing to Measurement Infrastructure for the End-to-end Performance Initiative
39
Active Measurements within Abilene
Surveyors with:Active delay/loss measurementsAd hoc throughput tests
42
Divide and Conquer
Systematically identify/isolate the network segment at fault
Can we make this systematic and (eventually) automated?
44
Why the End-to-End Performance Initiative?
Even with high bandwidth network links, the Internet2 community often does not see expected performance.
46
The E2Epi Mission
To enable the researchers, faculty, students and staff who use high performance networks to obtain optimal performance from the current infrastructure on a consistent basis.
Raw Connectivity
Applications Performance
47
True End-to-End Experience
•User perception
•Application
•Operating system
•Host IP stack
•Host network card
•Local Area Network
•Campus backbone network
•Campus link to regional network/GigaPoP
•GigaPoP link to Internet2 national backbones
•International connections
EYEBALL
APPLICATION
STACK
JACK
NETWORK
. . .
. . .
. . .
. . .
48
The Problem
Applications Developer
System Administrator
LAN Administrator
CampusNetworking
Gigapop Gigapop
Backbone
CampusNetworking
LAN Administrator
System Administrator
Applications Developer
How do you solvea problem along a path?
Hey, this is not working right!
The computerIs working OK
Talk to the other guys
Everything isAOK
No othercomplaints
The network is lightly loaded
All the lights are green
We don’t see anything wrong
Looks fine
Others are getting in ok
Not our problem
49
First Steps
Workshop in Ann Arbor on 9 January, 2001
• 40+ participants• Each participant provided a short paper on “What does E2EPerformance Mean?”
• Planned agenda was not used in order to respond to more pressing issues from participants.
• Design team formed to create an overall vision paper.
50
Areas of the Initiative
Applications
Host/OS Tuning
Measurement Infrastructure
Performance Improvement Environment (PIE)
Operations and Human Communications
Performance Evaluation and Review Framework (PERF)
51
Applications
•Work with specific application communities to help solve their performance problems.
• High Energy Physics• Medical Sciences – Visible Human Project
•Use a few key, general purpose applications for performance testing.
• FTP• Video Conferencing
52
Host/OS Tuning
•Web100 has a leading role
•Provide Best Practices for getting the most from your computer.
•Locate or build tools for Host/OS performance diagnostics.
•Work with OS vendors on tuning capabilities
•Work with computer vendors on Internet2 Performance Packages.
53
Measurement Infrastructure
•Bring together current measurement efforts and projects in the community.
•Establish an End-to-End Measurement Infrastructure from the intersection of these works.
•Create diagnostic tools to determine the health of the network and locate performance problems.
54
Standard Operational Info
Applications Developer
System Administrator
LAN Administrator
CampusNetworking
Gigapop Gigapop
Backbone
CampusNetworking
LAN Administrator
System Administrator
Applications Developer
Information fromAll Parts of the Network
Ops Info
Ops Info
Ops Info
Ops Info
Ops Info
Ops Info Ops Info
Ops InfoOps Info
55
Standard Operational Info
Applications Developer
System Administrator
LAN Administrator
CampusNetworking
Gigapop Gigapop
Backbone
CampusNetworking
LAN Administrator
System Administrator
Applications Developer
Know the Healthof the Network
Ops Info
Ops Info
Ops Info
Ops Info
Ops Info
Ops Info Ops Info
Ops InfoOps Info
End-to-End Analyzer
56
Standard Operational Info
Applications Developer
System Administrator
LAN Administrator
CampusNetworking
Gigapop Gigapop
Backbone
CampusNetworking
LAN Administrator
System Administrator
Applications Developer
ApplicationsAdapt to the Network
Ops Info
Ops Info
Ops Info
Ops Info
Ops Info
Ops Info Ops Info
Ops InfoOps Info
End-to-End AnalyzerApps Tuning
57
Performance Improvement Environment (PIE)
•Develop a dynamic environment where collaboration and information sharing will happen.
•Identify, collect and disseminate appropriate information for end-to-end related information.
•Include success stories,measurement statistics, reference materials, measurement tool documentation.
•Include pointers to materials already developed by other communities.
58
Operations and Human Communications
Establish communications among common interest groups
• System administrators• LAN administrators• Campus NOCs• GigaPoP• Application support staff
Establish communications between groups for operations and problem resolution.
59
Groups of Common Interest
Applications Developer
System Administrator
CampusNetworking
Gigapop Gigapop
Backbone
CampusNetworking
LAN Administrator
System Administrator
Applications Developer
Provide a means of communications
Let them shareexperiences.
I don’t know how to solve this problem! I do! I do!
LAN Administrator
60
Find a Solution?
Applications Developer
System Administrator
LAN Administrator
CampusNetworking
Gigapop Gigapop
Backbone
CampusNetworking
LAN Administrator
System Administrator
Applications Developer
A System to Check aSpecific Problem
Hmm, Time to Check the PIE and
Talk to others
Performance OK here
Throughput OK
Everything isStill AOK
Not a bottleneckAt this point
The network is still lightly loaded
All the lights are green
We don’t see anything wrong
Yup. DuplexDoes notAgree!
It is slow for others too!
Ah ha, an Ethernet Duplex
problem!
61
Can You Go Direct to the Problem?
How can you tell where is the problem?
Need a tool to tell you:• Where the problem is.• The type of problem• Who to contact to get it fixed
Terry Gray, University of Washington“We Need a Finger Pointing Tool”
62
Gray Finger Pointing Tool
Applications Developer
System Administrator
LAN Administrator
CampusNetworking
Gigapop Gigapop
Backbone
CampusNetworking
LAN Administrator
System Administrator
Applications Developer
Locate the ProblemGray Finger Pointing ToolTerry GrayUniversity of Washington
Finger Pointing Tool Time!
63
Gray Finger Pointing Tool
Applications Developer
System Administrator
LAN Administrator
CampusNetworking
Gigapop Gigapop
Backbone
CampusNetworking
LAN Administrator
System Administrator
Applications Developer
Finger Pointing Tool Time
You ARE the Weakest Link!
OK, I’ll fix it
64
Performance Evaluation and Review Framework (PERF)
Establish a framework for resolving performance problems
• Finger Pointing Tool• Provide known solutions by using the PIE• Tap community knowledge by facilitating group communications
• Coordinate a team of experts to solve hard problems
65
The Hard Problems
Applications Developer
System Administrator
LAN Administrator
CampusNetworking
Gigapop Gigapop
Backbone
CampusNetworking
LAN Administrator
System Administrator
Applications Developer
What if no one has the answer?
Hey, Its is not working again!
I don’t know what is wrong
No clue here
I’m stumped
This is strange
It looks normal here
Doh!
We can’t figure it out
I amCluefully
Challenged
Others are getting in ok
I don’t know
Need a Tiger Team of Experts to Research the Problem
66
Anticipated Partners
Faculty and discipline communities
Campuses
GigaPoPs
International partners
Research projects in performance
Internet2 corporate members
Federal labs and agencies
67
Calls For Participation
Identify core applications and services
Seek stories and best practices• Current Call for Experiences
Seek participants in the various work areas• Internet2 E2Epi Measurement WorkshopTempe, AZ, 27-28 January 2002
• Campus Participation in E2Epi
68
Internet2 Organization Role
Staffing• Cheryl Munn-Fremon, Initiative Director• Russ Hobby, Technical Architect• George Brett, Information Architect• Lisa Wilberding, Communications Coordinator• Terri Saarinen, Program Assistant
71
Advanced Network Management
Layer 2 QueryProtocolSteven Wallace
Mark Meiss
Indiana University Advanced Network Management Laboratory
72
Presentation Overview
Why is there a need for layer 2 visualization
Why use a proxy agent?
Why host the proxy in the router?
How does this work?
73
The Need for Layer 2 Visualization
Many end-to-end performance problems are caused by defects in the end-system’s broadcast domain• Frequently caused by duplex mismatches
Topology of broadcast domains typically not known
Hop by hop analysis requires you to know the hops
74
Why Use a Proxy Agent To Implement this Protocol?
Frequently network engineers from “other” organizations assist in network troubleshooting
Rather than open SNMP access to some other organization, develop a service and related protocol to allow a remote engineer to safely determine the broadcast domain topology and health
75
Why Host the Proxy in the Router?
The router is in the path of the probe (which is in the form of a traceroute probe), and will automatically route the packet to the general purpose CPU due to the expired TTL
The router “knows” something about the broadcast domain and is a good point of instrumentation.
76
How Does Layer 2 Query Protocol Work?
Assumes broadcast domain made up of SNMP managed switches that support the standard bridge MIB
Switches discovered via an IP broadcast SNMP query
ARP information retrieved for all of the switch SNMP agents by querying the router
Subsets of the Ethernet forwarding tables are retrieved from the switches SNMP agents
Topology is calculated
77
The Protocol
Initial request looks like a traceroute probe with the addition of a special signature and three arguments: client’s IP address, port number, and initial query sequence number
Router-based agent establishes a TCP connection to the client on the given port and sends the initial query sequence number
Client and agent exchange commands over the TCP connection. Responses from the router are encoded in XML
80
Internet2 Contact Information
Guy Almes: [email protected]
Jill Arnold: [email protected]
Steve Corbato: [email protected]
Ted Hanss: [email protected]
Russ Hobby: [email protected]
Doug Van Houweling: [email protected]
Steve Wallace: [email protected]