160 likes | 276 Views
Join us for an intensive workshop on advanced networking focused on end-to-end performance, functionality, and the engineering objectives of Internet2. This workshop, led by Paul Love from the Topology Working Group, will cover the current landscape of Internet2, the aspirations for improved networking capabilities, and the challenges such as bandwidth, quality of service, and security. Participants will gain insights on enhancing user experience, addressing technological threats, and supporting advanced network services. Aimed at researchers and educators interested in optimizing network performance.
E N D
Campus Focused Workshop on Advanced Networking Paul Love Chair, Topology Working Group Campus Workshop Houston 10-11 April 2002
Outline • Internet2 Engineering Objectives • Hopes for & Threats to End-to-End Performance • A few words on Abilene
Engineering Objectives of Internet2 • Provide our members with superlative networking • Performance • Functionality • Understanding • Make superlative networking strategic to research & education
End-to-End: Challenge, Aspirations & Threats • Support services of advanced networks E2E (eyeball2eyeball) • Performance • Current target: 80Mb/s across the country • Multiplies where possible • Functions • Multicast • IPv6 • Quality of Service • Measurement • Security
What are our Aspirations? • Switched 100BaseT + well-provisioned Internet2 networking @ 80 Mb/s (for now) • But user expectations and experiences vary widely • Don’t take the easy way out • Boost expectations & experiences - raise the bar • Raise the bar again – work hard to stay out there
Threats • Distance BW = C x packet-size / ( delay x sqrt(packet-loss ))(Mathis, Semke, Mahdavi, and Ott, CCR, July 1997) • Fiber: dirty connections, bad light/connectors • Switches: full/half duplex & 10/100 mismatches, head of line blocking • Routing: Asymmetric, increased distance • Provisioning: a “straw” somewhere • Host: OS & TCP stack, H/W, Apps
09 January 2002 Sacramento Washington Los Angeles Abilene International Peering STAR TAP/Star Light APAN/TransPAC, Ca*net3, CERN, CERnet, FASTnet, GEMnet, IUCC, KOREN/KREONET2, NORDUnet, RNP2, SURFnet, SingAREN, TAnet2 Pacific Wave AARNET, APAN/TransPAC, CA*net3, TANET2 NYCM BELNET, CA*net3, GEANT*, HEANET, JANET, NORDUnet SNVA GEMNET, SINET, SingAREN, WIDE LOSA UNINET OC3-OC12 San Diego (CALREN2) CUDI AMPATH REUNA, RNP2 RETINA (ANSP) El Paso (UACJ-UT El Paso) CUDI * ARNES, CARNET, CESnet, DFN, GRNET, RENATER, RESTENA, SWITCH, HUNGARNET, GARR-B, POL-34, RCCN, RedIRIS
Raw HDTV/IP testing • Packetized raw HDTV (1.5 Gbps) • ISIe, Tektronix, & UW project/DARPA support • Connectivity and testing support • P/NW & MAX Gigapops, Abilene and DARPA Supernet, Level(3) • SC2001 public demo • November, 2001 • SEA -> DEN via L(3) OC-48c SONET
Raw HDTV/IP Demo • DARPA PIs Meeting: SEA->DC area 1/6/02 • 18 hrs of continuous, single-stream raw HD/IP • UDP jumbo frames: 4444 B packet size • Application level measurement • 3 billion packets transmitted • 0 packets lost, 15 resequencing episodes • e2e network performance • Loss: <8x10 -10 (90% confidence level) • Reordering: 5x10 –9 • Transcontinental 1-Gbps TCP (std 1.5 kB MTU) requires loss at the level of 3x10 –8 or lower
Where things are at Present • Infrastructure of large capacity • Besides the HDTV/IP demos we have examples of 240Mb/s flows • But flows aren’t predictable – even 40Mb/s • People don’t know what they should expect
Why Care? • Faculty needs keep advancing: • Effective access to remote facility: quickly move large datasets. PPDG: 400 Mb/s to CERN by 2003 • Interactive access: video or control or VoIPVery low loss/jitter • We (in several senses) need to deliver • Low aspirations are dangerous to us, to our goals
Baseline BW Requirements for the US-CERN Transatlantic Link With thanks to Harvey B Newman, CIT