Olivier Martin, CERN NEC’2005 Conference, VARNA (Bulgaria)

42
16 September 2005 1 The ongoing evolution from Packet based networks to Hybrid Networks in Research & Education Networks Olivier Martin, CERN NEC’2005 Conference, VARNA (Bulgaria)

description

The ongoing evolution from Packet based networks to Hybrid Networks in Research & Education Networks. Olivier Martin, CERN NEC’2005 Conference, VARNA (Bulgaria). Presentation Outline. The demise of conventional packet based networks in the R&E community - PowerPoint PPT Presentation

Transcript of Olivier Martin, CERN NEC’2005 Conference, VARNA (Bulgaria)

  • The ongoing evolution from Packet based networks to Hybrid Networks in Research & Education NetworksOlivier Martin, CERNNEC2005 Conference, VARNA (Bulgaria)

  • Presentation Outline

    The demise of conventional packet based networks in the R&E communityThe advent of community managed dark fiber networksThe Grid & its associated Wide Area Networking challengeson-demand Lambda GridsEthernet over SONET & new standardsWAN-PHY, GFP, VCAT/LCAS, G.709, OTN

  • OC-768c40-GE

  • Internet Backbone SpeedsT1 LinesT3 linesOC3cOC12cIP/ATM-VCsMBPS

    Chart1

    0.056

    0.256

    4.5

    4.5

    10

    20

    40

    180

    360

    900

    2000

    4000

    8000

    100000

    5000000

    Backbone Speed

    Internet Backbone Speed (in Mbps)

    Sheet1

    YearBackbone Speed

    19860.05656,000

    19870.256256,000

    19984.51,560,000

    19894.51,560,000

    1990103,000,000

    1991204,500,000

    19924010,000,000

    199318045,000,000

    1994360130,000,000

    1995900300,000,000

    19962000900,000,000

    199740002,000,000,000

    199880004,000,000,000

    1999100000100,000,000,000

    200050000005,000,000,000,000

    Sheet1

    Backbone Speed

    Internet Backbone Speed (in Mbps)

    Sheet2

    Sheet3

  • High Speed IP Network Transport TrendsHigher Speed, Lower cost, complexity and overheadB-ISDNATMSONET/SDHIPOpticalMultiplexing, protection and management at every layerSignalling

  • Network ExponentialsNetwork vs. computer performanceComputer speed doubles every 18 monthsNetwork speed doubles every 9 monthsDifference = order of magnitude per 5 years1986 to 2000Computers: x 500Networks: x 340,0002001 to 2010Computers: x 60Networks: x 4000Moores Law vs. storage improvements vs. optical improvements. Graph from Scientific American (Jan-2001) by Cleo Vilett, source Vined Khoslan, Kleiner, Caufield and Perkins.

    l

    Know the userBW requirements# of usersCABA -> Lightweight users, browsing, mailing, home useB -> Business applications, multicast, streaming, VPNs, mostly LANC -> Special scientific applications, computing, data grids, virtual-presenceADSLGigE LAN(3 of 12)F(t)

    l

    What the userBW requirementsTotal BWCABA -> Need full Internet routing, one to manyB -> Need VPN services on/and full Internet routing, several to several C -> Need very fat pipes, limited multiple Virtual Organizations, few to fewADSLGigE LAN(4 of 12)

    l

    So what are the factsCosts of fat pipes (fibers) are one/third of equipment to light them upIs what Lambda salesmen told Cees de Laat (University of Amsterdam & Surfnet) Costs of optical equipment 10% of switching 10 % of full routing equipment for same throughput100 Byte packet @ 10 Gb/s -> 80 ns to look up in 100 Mbyte routing table (light speed from me to you on the back row!)Big sciences need fat pipesBottom line: create a hybrid architecture which serves all users in one coherent and cost effective way(5 of 12)

  • Utilization trends

    GbpsNetwork Capacity LimitJan 2005

  • Todays hierarchical IP networkUniversityNational or Pan-National IP NetworkOther national networksNREN ANREN BNREN CNREN D

  • Tomorrows peer to peer IP networkWorldUniversityServerWorldWorldNational DWDM NetworkNREN ANREN BNREN CNREN DChildLightpathsChild Lightpaths

  • Creation of application VPNsCommodityInternetBio-informaticsNetworkUniversityUniversityUniversityCERNUniversityUniversityHigh Energy Physics NetworkeVLBI NetworkDeptResearch NetworkDirect connect bypasses campus firewall

  • Production vs Research Campus NetworksIncreasingly campuses are deploying parallel networks for high end usersReduces costs by providing high end network capability to only those who need itLimitations of campus firewall and border router are eliminatedMany issues in regards to security, back door routing, etcCampus networks may follow same evolution as campus computingDiscipline specific networks being extended into the campus

  • UCLP intended for projects like National LambdaRail

  • GEANT2 POP Design

    EMBED Word.Picture.8

    _1147001627.doc

    Nx10Gbps to other GANT2 PoP

    2x10Gbps to local NREN

    Dark fibre to other GANT2 PoP

    GANT2 PoP

    Juniper M-160

    DWDM

  • UltraLight Optical Exchange PointL1, L2 and L3 servicesInterfaces1GE and 10GE10GE WAN-PHY (SONET friendly)Hybrid packet- and circuit-switched PoPInterface between packet- & circuit-switched networksControl plane is L3Photonic switchCalient or Glimmerglass Photonic Cross Connect Switch

  • LHC Data Grid HierarchyTier 1Online System CERN 700k SI95 ~1 PB Disk; Tape RobotFNAL: 200k SI95; 600 TBIN2P3 Center INFN Center RAL Center InstituteInstituteInstituteInstitute ~0.25TIPSWorkstations~100-400 MBytes/sec2.5/10 Gbps0.11 GbpsPhysicists work on analysis channelsEach institute has ~10 physicists working on one or more channelsPhysics data cache~PByte/sec10 Gbps~2.5 GbpsTier 0 +1Tier 3Tier 4Tier 2ExperimentCERN/Outside Resource Ratio ~1:2 Tier0/( Tier1)/( Tier2) ~1:1:1

  • Deploying the LHC [email protected] LHC Computing Centre

  • What you [email protected]

  • Main Networking Challenges

    Fulfill the, yet unproven, assertion that the network can be nearly transparent to the GridDeploy suitable Wide Area Network infrastructure (50-100 Gb/s)Deploy suitable Local Area Network infrastructure (matching or exceeding that of the WAN)Seamless interconnection of LAN & WAN infrastructuresfirewall?End to End issues (transport protocols, PCs (Itanium, Xeon), 10GigE NICs (Intel, S2io), where are we today:memory to memory: 7.5Gb/s (PCI bus limit)memory to disk: 1.2MB (Windows 2003 server/NewiSys)disk to disk: 400MB (Linux), 600MB (Windows)

  • Main TCP issuesDoes not scale to some environmentsHigh speed, high latencyNoisyUnfair behaviour with respect to:Round Trip Time (RTTFrame size (MSS)Access BandwidthWidespread use of multiple streams in order to compensate for inherent TCP/IP limitations (e.g. Gridftp, BBftp):Bandage rather than a cureNew TCP/IP proposals in order to restore performance in single stream environmentsNot clear if/when it will have a real impactIn the mean time there is an absolute requirement for backbones with:Zero packet losses,And no packet re-orderingWhich re-inforces the case for lambda Grids

  • TCP dynamics(10Gbps, 100ms RTT, 1500Bytes packets)Window size (W) = Bandwidth*Round Trip TimeWbits = 10Gbps*100ms = 1GbWpackets = 1Gb/(8*1500) = 83333 packetsStandard Additive Increase Multiplicative Decrease (AIMD) mechanisms:W=W/2 (halving the congestion window on loss event)W=W + 1 (increasing congestion window by one packet every RTT)Time to recover from W/2 to W (congestion avoidance) at 1 packet per RTT:RTT*Wp/2 = 1.157 hourIn practice, 1 packet per 2 RTT because of delayed acks, i.e. 2.31 hourPackets per second:RTT*Wpackets = 833333 packets

  • Single TCP stream performance under periodic lossesTCP throughput much more sensitive to packet loss in WANs than LANsTCPs congestion control algorithm (AIMD) is not well-suited to gigabit networksThe effect of packets loss can be disastrousTCP is inefficient in high bandwidth*delay networksThe future performance-outlook for computational grids looks bad if we continue to rely solely on the widely-deployed TCP RENOLoss rate =0.01%:LAN BW utilization= 99%WAN BW utilization=1.2%Bandwidth available = 1 Gbps

    Chart7

    1.265957446822.8723404255

    4.031914893635.4255319149

    12.765957446864.6808510638

    19.893617021377.7659574468

    27.872340425597.4468085106

    34.468085106497.9787234043

    39.468085106498.4042553191

    53.617021276698.5106382979

    0.00000498.9361702128

    0.000003333399.5744680851

    0.00000299.6808510638

    0.000001666799.7872340426

    0.0000013333100

    WAN (RTT=120ms)

    LAN (RTT=0.04 ms)

    Packet Loss frequency (%)

    Bandwidth Utilization (%)

    Effect of packet loss

    CHI

    1/PBW (MbpsBW (Mbpsbw utilsation50900150200250300350400450500550600650

    0

    10100.26781249780.37555315790.0023360.0001297778215

    10010.8468974791.18760336130.023360.0012977778333

    10000.12.67812497833.75553157870.23360.0129777778608

    100000.018.468974789911.876033613411.91.26595744682.3360.1297777778731

    1000000.00126.781249782737.555315787237.94.031914893623.361.2977777778

    10000000.000184.6897478992118.760336134512012.7659574468233.612.9777777778570776.255707763428082.191780822916

    25000000.00004133.9062489134187.77657893618719.893617021358432.4444444444921

    50000000.00002189.3720332998265.556184627426227.8723404255116864.8888888889925

    75000000.0000133333231.932426569325.238575188732434.4680851064175297.3333333333926

    100000000.00001267.8124978268375.553157872137139.46808510642336129.7777777778930

    175000000.0000057143354.2826336224496.810129907250453.61702127664088227.1111111111936

    250000000.000004423.4487394958593.80168067235840324.4444444444937

    300000000.0000033333463.864853138650.47715037747008389.3333333333938

    500000000.000002598.8469503648839.762390166711680648.8888888889940

    600000000.0000016667656.0039664159919.913608077514016778.6666666667940

    750000000.0000013333733.43473120771028.494680544217520973.3333333333940

    1000000000.000001846.89747899161187.6033613445233601297.7777777778

    10000000000.00000012678.1249782683755.531578720723360012977.7777777778

    CHI

    00215

    00333

    00608

    00731

    00916

    00921

    00925

    00926

    00930

    00936

    00937

    00938

    00940

    00940

    00940

    00940

    00940

    00940

    MSS*C/ (RTT*sqrt(p))

    WAN (RTT=120ms)

    LAN (RTT=0.04 ms)

    Packet Loss frequency (one packet every x packets transmitted)

    Throuhput (Mbps)

    Effect of packet loss

    Gva

    022.8723404255

    035.4255319149

    064.6808510638

    077.7659574468

    097.4468085106

    097.9787234043

    098.4042553191

    098.5106382979

    098.9361702128

    099.5744680851

    099.6808510638

    099.7872340426

    0100

    0100

    0100

    WAN (RTT=120ms)

    LAN (RTT=0.04 ms)

    Packet Loss frequency (one packet every x packets transmitted)

    Bandwidth Utilisation %

    Effect of packet loss

    Sheet3

    00215

    00333

    00608

    00731

    00916

    00921

    00925

    00926

    00930

    00936

    00937

    00938

    00940

    00940

    00940

    940

    940

    940

    MSS*C/ (RTT*sqrt(p))

    WAN (RTT=120ms)

    LAN (RTT=0.04 ms)

    Packet Loss frequency (%)

    Throuhput (Mbps)

    Effect of packet loss

    022.8723404255

    035.4255319149

    064.6808510638

    077.7659574468

    097.4468085106

    097.9787234043

    098.4042553191

    098.5106382979

    098.9361702128

    099.5744680851

    099.6808510638

    099.7872340426

    0100

    WAN (RTT=120ms)

    LAN (RTT=0.04 ms)

    Packet Loss frequency (%)

    Bandwidth Utilisation (%)

    Effect of packet loss

    w05gva -> w06gva

    w=1M

    RTT=

    txqueuelen = 100

    Loss rate (%)Rate (Mbit/s)Link Utilisation (%)

    10101021522.87234042550.0005432558

    205533335.42553191490.0007015015

    502260864.68085106380.0009605263

    1001173177.76595744680.0015978112

    5000.20.291697.44680851060.0063755459

    7500.13333333330.133333333392197.97872340430.0095114007

    10000.10.192598.40425531910.012627027

    50000.020.0292698.51063829790.0630669546

    100000.010.0193098.93617021280.1255913978

    1000000.0010.00193699.57446808511.2478632479939

    10000000.00010.000193799.680851063812.4653148346

    25000000.000040.0000493899.787234042631.1300639659

    50000000.000020.0000294010062.1276595745

    75000000.00001333330.000013333394010093.1914893617

    100000000.000010.00001940100124.2553191489

    175000000.00000571430.0000057143940100217.4468085106

    250000000.0000040.000004940100310.6382978723

    300000000.00000333330.0000033333940100372.7659574468

    09

  • ResponsivenessLarge MTU accelerates the growth of the windowTime to recover from a packet loss decreases with large MTULarger MTU reduces overhead per frames (saves CPU cycles, reduces the number of packets) r =C . RTT2 . MSS2C : Capacity of the linkTime to recover from a single packet loss:

  • Internet2 land speed record history (IPv4 & IPv6) period 2000-2004

    Chart1

    MonthMonthMonthMonth

    0.760.956Mar-00Mar-00

    0.4020.402Apr-02Apr-02

    Sep-02Sep-020.4830.483

    Oct-02Oct-020.3480.348

    0.9230.923Nov-02Nov-02

    2.382.38Feb-03Feb-03

    May-03May-030.9830.983

    5.445.44Oct-03Oct-03

    5.645.64Nov-03Nov-03

    Nov-03Nov-0344

    Feb-046.25Feb-04Feb-04

    4.2226Apr-04Apr-04Apr-04

    May-047.09May-04May-04

    IPv4 (Gb/s) single stream

    IPv4 (Gb/s) multiple streams

    IPv6 (Gb/s) single stream

    IPv6 (Gb/s) multiple streams

    DataTAG-Web-Hits

    4241

    4910

    3889

    6949

    8788

    17356

    18616

    20420

    18851

    37266

    36076

    Hits/month

    Month

    Hits

    DataTAG Web site

    DataTAG-Web-Hits-v2

    Month

    4241

    4910

    3889

    6949

    8788

    17356

    18616

    20420

    18851

    37266

    36076

    31379

    25998

    23407

    26423

    28340

    22866

    27250

    40518

    29530

    28480

    30335

    58530

    79073

    DataTAG

    Month

    Hits

    IPv4-IPv6-Gbs

    IPv4 (Gb/s) multiple streams

    IPv6 (Gb/s) multiple streams

    Month

    Internet2 landspeed record history(in Gigabit/second)

    IPv4-IPv6-Tbms

    MonthMonth

    4278Mar-00

    4933Apr-02

    Sep-021215

    Oct-025154

    10136Nov-02

    23886Feb-03

    May-036947

    38420Oct-03

    61752Nov-03

    Nov-0346156

    IPv4 terabit-meters/second)

    IPv6 (terabit-meters/second)

    Month

    Internet2 landspeed record history(in terabit-meters/second)

    IP-v4-IPv6-Gbs-v1

    IPv4 (Gb/s) multiple streams

    IPv6 (Gb/s) multiple streams

    Month

    Evolution of the I2LSR in Gigabit/second

    IPv4-IPv6-Tbms-v1

    MonthMonth

    4278Mar-00

    4933Apr-02

    Sep-021215

    Oct-025154

    10136Nov-02

    23886Feb-03

    May-036947

    38420Oct-03

    61752Nov-03

    Nov-0346156

    IPv4 terabit-meters/second)

    IPv6 (terabit-meters/second)

    Month

    Evolution of the I2LSR using terabit-meters/second metrics

    IPv4-v6-Tbms

    MonthMonthMonthMonthMonthMonth

    Mar-00Mar-00Mar-00Mar-004278Mar-00

    Apr-02Apr-02Apr-02Apr-024933Apr-02

    Sep-02Sep-02Sep-02Sep-02Sep-021215

    Oct-02Oct-02Oct-02Oct-02Oct-025154

    Nov-02Nov-02Nov-02Nov-0210136Nov-02

    Feb-03Feb-03Feb-03Feb-0323886Feb-03

    May-03May-03May-03May-03May-036947

    Oct-03Oct-03Oct-03Oct-0338420Oct-03

    Nov-03Nov-03Nov-03Nov-0361752Nov-03

    Nov-03Nov-03Nov-03Nov-03Nov-0346156

    Feb-04Feb-04Feb-04Feb-0468431Feb-04

    Apr-04Apr-04Apr-04Apr-0469073Apr-04

    May-04May-04May-04May-0477699May-04

    IPv4 (Gb/s) single stream

    IPv4 (Gb/s) multiple streams

    IPv6 (Gb/s) single stream

    IPv6 (Gb/s) multiple streams

    IPv4 terabit-meters/second)

    IPv6 (terabit-meters/second)

    MONTH

    TERABIT-METERS/SECOND

    EVOLUTION OF THE I2LSR IN TERABITS-METERS/SECOND

    IPv4-v6-Gbs

    MonthMonthMonthMonth

    0.760.956Mar-00Mar-00

    0.4020.402Apr-02Apr-02

    Sep-02Sep-020.4830.483

    Oct-02Oct-020.3480.348

    0.9230.923Nov-02Nov-02

    2.382.38Feb-03Feb-03

    May-03May-030.9830.983

    5.445.44Oct-03Oct-03

    5.645.64Nov-03Nov-03

    Nov-03Nov-0344

    Feb-046.25Feb-04Feb-04

    4.2226Apr-04Apr-04Apr-04

    May-047.09May-04May-04

    IPv4 (Gb/s) single stream

    IPv4 (Gb/s) multiple streams

    IPv6 (Gb/s) single stream

    IPv6 (Gb/s) multiple streams

    MONTH

    Gigabits/second

    Category

    EVOLUTION OF THE I2LSR IN GIGABITS/SECOND

    IPv4-IPv6

    IPv4 (Gb/s) single streamIPv4 (Gb/s) multiple streamsIPv6 (Gb/s) single streamIPv6 (Gb/s) multiple streamsIPv4 terabit-meters/second)IPv6 (terabit-meters/second)Owner(s)

    Month

    Mar-000.7600.9564,278Microsoft, Qwest, University of Washington, USC ISI

    Apr-020.4020.4024,933University of Alberta, University of Amsterdam (UvA), Surfnet

    Sep-020.4830.4831,215ARNES, DANTE, RedIris

    Oct-020.3480.3485,154ARNES, DANTE, RedIris

    Nov-020.9230.92310,136Caltech, DataTAG, Nikhef, SLAC, UvA

    Feb-032.382.3823,886Caltech, CERN, DataTAG, Los Alamos National Laboratory (LANL), SLAC

    May-030.9830.9836,947Caltech, CERN, DataTAG

    Oct-035.445.4438,420Caltech, CERN, DataTAG

    Nov-035.645.6461,752Caltech, CERN, DataTAG

    Nov-034446,156Caltech, CERN, DataTAG

    Feb-046.2568,431Caltech, CERN, DataTAG (multi-stream)

    Apr-044.222669,073SUNET

    May-047.0977,699Caltech, CERN, DataTAG (multi-stream)

    IPv4-IPv6

    IPv4 (Gb/s) single stream

    IPv4 (Gb/s) multiple streams

    IPv6 (Gb/s) single stream

    IPv6 (Gb/s) multiple streams

    Datag related Web sites

    MonthMonthMonth

    4241Apr-02Apr-02

    4910May-02May-02

    3889Jun-02Jun-02

    6949Jul-02Jul-02

    8788Aug-02Aug-02

    17356Sep-02Sep-02

    18616Oct-02Oct-02

    20420Nov-02Nov-02

    18851Dec-02Dec-02

    37266Jan-03Jan-03

    36076Feb-03Feb-03

    31379Mar-03Mar-03

    25998Apr-03Apr-03

    23407May-03May-03

    26423Jun-03Jun-03

    28340Jul-03Jul-03

    22866Aug-03Aug-03

    27250Sep-03Sep-03

    405181058Oct-03

    29530297Nov-03

    2848037916569

    303353351968

    58530306915

    DataTAG

    Telecom

    ICT4D

    Month

    Hits/month

    DataTAG related Web sites

    DataTAG Web site activity

    Month

    4241

    4910

    3889

    6949

    8788

    17356

    18616

    20420

    18851

    37266

    36076

    31379

    25998

    23407

    26423

    28340

    22866

    27250

    40518

    29530

    28480

    30335

    58530

    79073

    DataTAG

    Month

    Hits/month

    DataTAG Web site

    DataTAG Tables

    DataTAGGridcafeTelecomICT4D

    Month

    Apr-024241

    May-024910

    Jun-023889

    Jul-026949

    Aug-028788

    Sep-0217356

    Oct-0218616

    Nov-0220420

    Dec-0218851

    Jan-0337266

    Feb-0336076

    Mar-0331379

    Apr-0325998

    May-0323407

    Jun-0326423

    Jul-0328340

    Aug-0322866

    Sep-0327250

    Oct-03405181716761058

    Nov-0329530379547297

    Dec-032848031876437916569

    Jan-04303352700003351968

    Feb-0458530291255306915

    Mar-04790732699846001339

    Apr-0446165497284332818

    May-04409906141719-May

    Jun-04

    Sheet1

    year16002000250030000.800.710.630.5

    19984.05.06.08.0400.00400.00400.00400.00

    199920.025.031.038.080.0080.0080.0080.00

    200044.455.669.483.336.0036.0036.0036.00

    2001160.0200.0250.0300.010.0010.0010.0010.00Based on actual 155Mbps pri

    20021012.71265.81582.31898.71.581.581.581.58Based on actual 622Mbps pri

    20031265.81582.31977.82373.41.261.121.000.79

    20041582.31977.82472.32966.81.010.800.630.40

    20051977.82472.33090.43708.50.810.570.400.20

    20062472.33090.43863.04635.60.650.400.250.10

    20073090.43863.04828.75794.50.520.290.160.05

    20083863.04828.76035.97243.10.410.200.100.02

    year1600.02000.02500.03000.0

    19984.05.06.08.0

    199920.025.031.038.0

    200044.455.669.483.3

    2001160.0200.0250.0300.0

    20021012.71265.81582.31898.7

    20031426.31782.82228.62674.3

    20042008.82511.13138.83766.6

    20052829.43536.74420.95305.0

    20063985.04981.36226.67471.9

    20075612.77015.98769.810523.8

    20087905.29881.512351.914822.2

    year1600.02000.02500.03000.0

    19984.05.06.08.0

    199920.025.031.038.0

    200044.455.669.483.3

    2001160.0200.0250.0300.0

    20021012.71265.81582.31898.7

    20031607.42009.22511.63013.9

    20042551.43189.33986.64783.9

    20054049.95062.36327.97593.5

    20066428.48035.510044.312053.2

    200710203.812754.715943.419132.0

    200816196.420245.625306.930368.3

    year1600200025003000

    19984.05.06.08.0

    199920.025.031.038.0

    200044.455.669.483.3

    2001160.0200.0250.0300.0

    20021012.71265.81582.31898.7

    20032025.32531.63164.63797.5

    20044050.65063.36329.17594.9

    20058101.310126.612658.215189.9

    200616202.520253.225316.530379.7

    200732405.140506.350632.960759.5

    200864810.181012.7101265.8121519.0

    622 Kb[s-0.20-0.29-0.37-0.50Hypothesis 622Mbps=2*155Mbps

    20013100.003100.003100.003100.00

    2002982.76982.76982.76982.76

    2003786.21697.76619.14491.38

    2004628.97495.41390.06245.69

    2005503.17351.74245.74122.85

    2006402.54249.74154.8161.42

    2007322.03177.3197.5330.71

    2.5 Gb[s-0.20-0.29-0.33-0.50Hypothesis 2.5Gbps=2*622Mbps

    20016200.006200.006200.006200.00

    20021965.521965.521965.521965.52

    20031572.421395.521238.28982.76

    20041257.93990.82780.11491.38

    20051006.35703.48491.47245.69

    2006805.08499.47309.63122.85

    2007644.06354.62195.0761.42

    1 Gb[s-0.20-0.29-0.33-0.50EuropeUSA

    200110000.0010000.0010000.0010000.00

    20021580.001580.001580.001580.00

    20031264.001121.80995.40790.00

    20041011.20796.48627.10395.00376.26156.78206.9414.4966992465

    2005808.96565.50395.07197.50237.0498.77130.3753.6914786909

    2006647.17401.50248.9098.75149.3462.2282.1466.9621616327

    2007517.73285.07156.8049.3894.0839.2051.7585.0313163589

    2008414.19202.4098.7924.6959.2724.7032.60104.2952653753

    Sheet2

    0.800.710.630.5

    400.00400.00400.00400.001998

    80.0080.0080.0080.001999

    36.0036.0036.0036.002000

    10.0010.0010.0010.002001

    8.007.106.305.002002

    6.405.043.972.502003

    5.123.582.501.252004

    4.102.541.580.632005

    3.281.800.990.312006

    2.621.280.630.162007

    22522522522520012250

    36340846058020022900

    578734932148020033700

    87912571800360020044500

    DataTAG-Web-Hits

    4241

    4910

    3889

    6949

    8788

    17356

    18616

    20420

    18851

    37266

    36076

    Hits/month

    Month

    Hits

    DataTAG Web site

    DataTAG-Web-Hits-v2

    Month

    4241

    4910

    3889

    6949

    8788

    17356

    18616

    20420

    18851

    37266

    36076

    31379

    25998

    23407

    26423

    28340

    22866

    27250

    40518

    29530

    28480

    30335

    58530

    79073

    DataTAG

    Month

    Hits

    IPv4-IPv6-Gbs

    IPv4 (Gb/s) multiple streams

    IPv6 (Gb/s) multiple streams

    Month

    Internet2 landspeed record history(in Gigabit/second)

    IPv4-IPv6-Tbms

    MonthMonth

    4278Mar-00

    4933Apr-02

    Sep-021215

    Oct-025154

    10136Nov-02

    23886Feb-03

    May-036947

    38420Oct-03

    61752Nov-03

    Nov-0346156

    IPv4 terabit-meters/second)

    IPv6 (terabit-meters/second)

    Month

    Internet2 landspeed record history(in terabit-meters/second)

    IP-v4-IPv6-Gbs-v1

    IPv4 (Gb/s) multiple streams

    IPv6 (Gb/s) multiple streams

    Month

    Evolution of the I2LSR in Gigabit/second

    IPv4-IPv6-Tbms-v1

    MonthMonth

    4278Mar-00

    4933Apr-02

    Sep-021215

    Oct-025154

    10136Nov-02

    23886Feb-03

    May-036947

    38420Oct-03

    61752Nov-03

    Nov-0346156

    IPv4 terabit-meters/second)

    IPv6 (terabit-meters/second)

    Month

    Evolution of the I2LSR using terabit-meters/second metrics

    IPv4-v6-Tbms

    MonthMonthMonthMonthMonthMonth

    Mar-00Mar-00Mar-00Mar-004278Mar-00

    Apr-02Apr-02Apr-02Apr-024933Apr-02

    Sep-02Sep-02Sep-02Sep-02Sep-021215

    Oct-02Oct-02Oct-02Oct-02Oct-025154

    Nov-02Nov-02Nov-02Nov-0210136Nov-02

    Feb-03Feb-03Feb-03Feb-0323886Feb-03

    May-03May-03May-03May-03May-036947

    Oct-03Oct-03Oct-03Oct-0338420Oct-03

    Nov-03Nov-03Nov-03Nov-0361752Nov-03

    Nov-03Nov-03Nov-03Nov-03Nov-0346156

    Feb-04Feb-04Feb-04Feb-0468431Feb-04

    Apr-04Apr-04Apr-04Apr-0469073Apr-04

    May-04May-04May-04May-0477699May-04

    IPv4 (Gb/s) single stream

    IPv4 (Gb/s) multiple streams

    IPv6 (Gb/s) single stream

    IPv6 (Gb/s) multiple streams

    IPv4 terabit-meters/second)

    IPv6 (terabit-meters/second)

    MONTH

    TERABIT-METERS/SECOND

    EVOLUTION OF THE I2LSR IN TERABITS-METERS/SECOND

    IPv4-v6-Gbs

    MonthMonthMonthMonth

    0.760.956Mar-00Mar-00

    0.4020.402Apr-02Apr-02

    Sep-02Sep-020.4830.483

    Oct-02Oct-020.3480.348

    0.9230.923Nov-02Nov-02

    2.382.38Feb-03Feb-03

    May-03May-030.9830.983

    5.445.44Oct-03Oct-03

    5.645.64Nov-03Nov-03

    Nov-03Nov-0344

    Feb-046.25Feb-04Feb-04

    4.2226Apr-04Apr-04Apr-04

    May-047.09May-04May-04

    IPv4 (Gb/s) single stream

    IPv4 (Gb/s) multiple streams

    IPv6 (Gb/s) single stream

    IPv6 (Gb/s) multiple streams

    MONTH

    Gigabits/second

    Category

    EVOLUTION OF THE I2LSR IN GIGABITS/SECOND

    Chart3

    IPv4 (Gb/s) single stream0.760.402IPv4 (Gb/s) single streamIPv4 (Gb/s) single stream0.9232.38IPv4 (Gb/s) single stream5.445.64IPv4 (Gb/s) single streamIPv4 (Gb/s) single stream4.2226IPv4 (Gb/s) single stream

    IPv4 (Gb/s) multiple streams0.9560.402IPv4 (Gb/s) multiple streamsIPv4 (Gb/s) multiple streams0.9232.38IPv4 (Gb/s) multiple streams5.445.64IPv4 (Gb/s) multiple streams6.25IPv4 (Gb/s) multiple streams7.09

    IPv6 (Gb/s) single streamIPv6 (Gb/s) single streamIPv6 (Gb/s) single stream0.4830.348IPv6 (Gb/s) single streamIPv6 (Gb/s) single stream0.983IPv6 (Gb/s) single streamIPv6 (Gb/s) single stream4IPv6 (Gb/s) single streamIPv6 (Gb/s) single streamIPv6 (Gb/s) single stream

    IPv6 (Gb/s) multiple streamsIPv6 (Gb/s) multiple streamsIPv6 (Gb/s) multiple streams0.4830.348IPv6 (Gb/s) multiple streamsIPv6 (Gb/s) multiple streams0.983IPv6 (Gb/s) multiple streamsIPv6 (Gb/s) multiple streams4IPv6 (Gb/s) multiple streamsIPv6 (Gb/s) multiple streamsIPv6 (Gb/s) multiple streams

    Month

    Mar-00

    Apr-02

    Sep-02

    Oct-02

    Nov-02

    Feb-03

    May-03

    Oct-03

    Nov-03

    Nov-03

    Feb-04

    Apr-04

    May-04

    Type

    Gigabit/second

    Month

    Evolution of Internet2 Landspeed record

    IPv4-IPv6

    IPv4 (Gb/s) single streamIPv4 (Gb/s) multiple streamsIPv6 (Gb/s) single streamIPv6 (Gb/s) multiple streamsIPv4 terabit-meters/second)IPv6 (terabit-meters/second)Owner(s)

    Month

    Mar-000.7600.9564,278Microsoft, Qwest, University of Washington, USC ISI

    Apr-020.4020.4024,933University of Alberta, University of Amsterdam (UvA), Surfnet

    Sep-020.4830.4831,215ARNES, DANTE, RedIris

    Oct-020.3480.3485,154ARNES, DANTE, RedIris

    Nov-020.9230.92310,136Caltech, DataTAG, Nikhef, SLAC, UvA

    Feb-032.382.3823,886Caltech, CERN, DataTAG, Los Alamos National Laboratory (LANL), SLAC

    May-030.9830.9836,947Caltech, CERN, DataTAG

    Oct-035.445.4438,420Caltech, CERN, DataTAG

    Nov-035.645.6461,752Caltech, CERN, DataTAG

    Nov-034446,156Caltech, CERN, DataTAG

    Feb-046.2568,431Caltech, CERN, DataTAG (multi-stream)

    Apr-044.222669,073SUNET

    May-047.0977,699Caltech, CERN, DataTAG (multi-stream)

    Datag related Web sites

    MonthMonthMonth

    4241Apr-02Apr-02

    4910May-02May-02

    3889Jun-02Jun-02

    6949Jul-02Jul-02

    8788Aug-02Aug-02

    17356Sep-02Sep-02

    18616Oct-02Oct-02

    20420Nov-02Nov-02

    18851Dec-02Dec-02

    37266Jan-03Jan-03

    36076Feb-03Feb-03

    31379Mar-03Mar-03

    25998Apr-03Apr-03

    23407May-03May-03

    26423Jun-03Jun-03

    28340Jul-03Jul-03

    22866Aug-03Aug-03

    27250Sep-03Sep-03

    405181058Oct-03

    29530297Nov-03

    2848037916569

    303353351968

    58530306915

    DataTAG

    Telecom

    ICT4D

    Month

    Hits/month

    DataTAG related Web sites

    DataTAG Web site activity

    Month

    4241

    4910

    3889

    6949

    8788

    17356

    18616

    20420

    18851

    37266

    36076

    31379

    25998

    23407

    26423

    28340

    22866

    27250

    40518

    29530

    28480

    30335

    58530

    79073

    DataTAG

    Month

    Hits/month

    DataTAG Web site

    DataTAG Tables

    DataTAGGridcafeTelecomICT4D

    Month

    Apr-024241

    May-024910

    Jun-023889

    Jul-026949

    Aug-028788

    Sep-0217356

    Oct-0218616

    Nov-0220420

    Dec-0218851

    Jan-0337266

    Feb-0336076

    Mar-0331379

    Apr-0325998

    May-0323407

    Jun-0326423

    Jul-0328340

    Aug-0322866

    Sep-0327250

    Oct-03405181716761058

    Nov-0329530379547297

    Dec-032848031876437916569

    Jan-04303352700003351968

    Feb-0458530291255306915

    Mar-04790732699846001339

    Apr-0446165497284332818

    May-04409906141719-May

    Jun-04

    Sheet1

    year16002000250030000.800.710.630.5

    19984.05.06.08.0400.00400.00400.00400.00

    199920.025.031.038.080.0080.0080.0080.00

    200044.455.669.483.336.0036.0036.0036.00

    2001160.0200.0250.0300.010.0010.0010.0010.00Based on actual 155Mbps pri

    20021012.71265.81582.31898.71.581.581.581.58Based on actual 622Mbps pri

    20031265.81582.31977.82373.41.261.121.000.79

    20041582.31977.82472.32966.81.010.800.630.40

    20051977.82472.33090.43708.50.810.570.400.20

    20062472.33090.43863.04635.60.650.400.250.10

    20073090.43863.04828.75794.50.520.290.160.05

    20083863.04828.76035.97243.10.410.200.100.02

    year1600.02000.02500.03000.0

    19984.05.06.08.0

    199920.025.031.038.0

    200044.455.669.483.3

    2001160.0200.0250.0300.0

    20021012.71265.81582.31898.7

    20031426.31782.82228.62674.3

    20042008.82511.13138.83766.6

    20052829.43536.74420.95305.0

    20063985.04981.36226.67471.9

    20075612.77015.98769.810523.8

    20087905.29881.512351.914822.2

    year1600.02000.02500.03000.0

    19984.05.06.08.0

    199920.025.031.038.0

    200044.455.669.483.3

    2001160.0200.0250.0300.0

    20021012.71265.81582.31898.7

    20031607.42009.22511.63013.9

    20042551.43189.33986.64783.9

    20054049.95062.36327.97593.5

    20066428.48035.510044.312053.2

    200710203.812754.715943.419132.0

    200816196.420245.625306.930368.3

    year1600200025003000

    19984.05.06.08.0

    199920.025.031.038.0

    200044.455.669.483.3

    2001160.0200.0250.0300.0

    20021012.71265.81582.31898.7

    20032025.32531.63164.63797.5

    20044050.65063.36329.17594.9

    20058101.310126.612658.215189.9

    200616202.520253.225316.530379.7

    200732405.140506.350632.960759.5

    200864810.181012.7101265.8121519.0

    622 Kb[s-0.20-0.29-0.37-0.50Hypothesis 622Mbps=2*155Mbps

    20013100.003100.003100.003100.00

    2002982.76982.76982.76982.76

    2003786.21697.76619.14491.38

    2004628.97495.41390.06245.69

    2005503.17351.74245.74122.85

    2006402.54249.74154.8161.42

    2007322.03177.3197.5330.71

    2.5 Gb[s-0.20-0.29-0.33-0.50Hypothesis 2.5Gbps=2*622Mbps

    20016200.006200.006200.006200.00

    20021965.521965.521965.521965.52

    20031572.421395.521238.28982.76

    20041257.93990.82780.11491.38

    20051006.35703.48491.47245.69

    2006805.08499.47309.63122.85

    2007644.06354.62195.0761.42

    1 Gb[s-0.20-0.29-0.33-0.50EuropeUSA

    200110000.0010000.0010000.0010000.00

    20021580.001580.001580.001580.00

    20031264.001121.80995.40790.00

    20041011.20796.48627.10395.00376.26156.78206.9414.4966992465

    2005808.96565.50395.07197.50237.0498.77130.3753.6914786909

    2006647.17401.50248.9098.75149.3462.2282.1466.9621616327

    2007517.73285.07156.8049.3894.0839.2051.7585.0313163589

    2008414.19202.4098.7924.6959.2724.7032.60104.2952653753

    Sheet2

    0.800.710.630.5

    400.00400.00400.00400.001998

    80.0080.0080.0080.001999

    36.0036.0036.0036.002000

    10.0010.0010.0010.002001

    8.007.106.305.002002

    6.405.043.972.502003

    5.123.582.501.252004

    4.102.541.580.632005

    3.281.800.990.312006

    2.621.280.630.162007

    22522522522520012250

    36340846058020022900

    578734932148020033700

    87912571800360020044500

  • Layer1/2/3 networking (1) Conventional layer 3 technology is no longer fashionable because of:High associated costs, e.g. 200/300 KUSD for a 10G router interfaces Implied use of shared backbonesThe use of layer 1 or layer 2 technology is very attractive because it helps to solve a number of problems, e.g. 1500 bytes Ethernet frame size (layer1)Protocol transparency (layer1 & layer2)Minimum functionality hence, in theory, much lower costs (layer1&2)

  • Layer1/2/3 networking (2) 0n-demand Lambda Grids are becoming very popular: Pros: circuit oriented model like the telephone network, hence no need for complex transport protocols Lower equipment costs (i.e. in theory a factor 2 or 3 per layer)the concept of a dedicated end to end light path is very elegant Cons: End to end still very loosely defined, i.e. site to site, cluster to cluster or really host to hostHigher circuit costs, Scalability, Additional middleware to deal with circuit set up/tear down, etcExtending dynamic VLAN functionality is a potential nightmare!

  • Lambda Grids What does it mean?

    Clearly different things to different people, hence the apparently easy consensus!Conservatively, on demand site to site connectivityWhere is the innovation?What does it solve in terms of transport protocols?Where are the savings?Less interfaces needed (customer) but more standby/idle circuits needed (provider)Economics from the service provider vs the customer perspective?Traditionally, switched services have been very expensive, Usage vs flat chargeBreak even, switches vs leased, few hours/dayWhy would this change? In case there are no savings, why bother?More advanced, cluster to clusterImplies even more active circuits in paralleIs it realistic?Even more advanced, Host to Host or even per flowAll opticalIs it really realisitic?

  • Some ChallengesReal bandwidth estimates given the chaotic nature of the requirements.End-end performance given the whole chain involved (disk-bus-memory-bus-network-bus-memory-bus-disk)Provisioning over complex network infrastructures (GEANT, NRENs etc)Cost model for options (packet+SLAs, circuit switched etc)Consistent Performance (dealing with firewalls)Merging leading edge research with production networking

  • Tentative conclusions

    There is a very clear trend towards community managed dark fiber networksAs a consequence National Research & Education Networks are evolving into Telecom Operators, is it right?In the short term, almost certainly YESIn the longer term, probably NOIn many countries, there is NO other way to have affordable access to multi-Gbit/s networks, therefore this is clearly the right moveThe Grid & its associated Wide Area Networking challengeson-demand Lambda Grids are, according to me, extremely doubtful!Ethernet over SONET & new standards will revolutionize the InternetWAN-PHY (IEEE) has, according to me NO future!However, GFP, VCAT/LCAS, G.709, OTN are very likely to have a very bright future.

  • Single TCP stream between Caltech and CERNAvailable (PCI-X) Bandwidth=8.5 Gbps RTT=250ms (16000 km) 9000 Byte MTU15 min to increase throughput from 3 to 6 Gbps

    Sending station: Tyan S2882 motherboard, 2x Opteron 2.4 GHz , 2 GB DDR.Receiving station: CERN OpenLab:HP rx4640, 4x 1.5GHz Itanium-2, zx1 chipset, 8GB memoryNetwork adapter: S2IO 10 GbE

    Burst of packet lossesSingle packet lossCPU load = 100%

  • High Throughput Disk to Disk Transfers: From 0.1 to 1GByte/sec Server Hardware (Rather than Network) Bottlenecks:Write/read and transmit tasks share the same limited resources: CPU, PCI-X bus, memory, IO chipset PCI-X bus bandwidth: 8.5 Gbps [133MHz x 64 bit]Link aggregation (802.3ad): Logical interface with two physical interfaces on two independent PCI-X buses.LAN test: 11.1 Gbps (memory to memory)

    Performance in this range (from 100 MByte/sec up to 1 GByte/sec) is required to build a responsive Grid-based Processing and Analysis System for LHC

  • Transferring a TB from Caltech to CERN in 64-bit MS Windows Latest disk to disk over 10Gbps WAN: 4.3 Gbits/sec (536 MB/sec) - 8 TCP streams from CERN to Caltech; 1TB file 3 Supermicro Marvell SATA disk controllers + 24 SATA 7200rpm SATA disksLocal Disk IO 9.6 Gbits/sec (1.2 GBytes/sec read/write, with
  • UltraLight: Developing Advanced Network Services for Data Intensive HEP ApplicationsUltraLight: a next-generation hybrid packet- and circuit-switched network infrastructurePacket switched: cost effective solution; requires ultrascale protocols to share 10G efficiently and fairlyCircuit-switched: Scheduled or sudden overflow demands handled by provisioning additional wavelengths; Use path diversity, e.g. across the US, Atlantic, Canada,Extend and augment existing grid computing infrastructures (currently focused on CPU/storage) to include the network as an integral componentUsing MonALISA to monitor and manage global systemsPartners: Caltech, UF, FIU, UMich, SLAC, FNAL, MIT/Haystack; CERN, NLR, CENIC, Internet2; Translight, UKLight, Netherlight; UvA, UCL, KEK, TaiwanStrong support from Cisco

  • UltraLight MPLS NetworkCompute path from one given node to another such that the path does not violate any constraints (bandwidth/administrative requirements)Ability to set the path the traffic will take through the network (with simple configuration, management, and provisioning mechanisms)Take advantage of the multiplicity of waves/L2 channels across the US (NLR, HOPI, Ultranet and Abilene/ESnet MPLS services) EoMPLS will be used to build layer2 pathsnatural step toward the deployment of GMPLS

  • SummaryFor many years the Wide Area Network has been the bottleneck; this is no longer the case in many countries thus making deployment of a data intensive Grid infrastructure possible!Recent I2LSR records show for the first time ever that the network can be truly transparent and that throughputs are limited by the end hostsChallenge shifted from getting adequate bandwidth to deploying adequate infrastructure to make effective use of it!Some transport protocol issues still need to be resolved; however there are many encouraging signs that practical solutions may now be in sight.1GByte/sec disk to disk challenge. Today: 1 TB at 536 MB/sec from CERN to Caltech Still in Early Stages; Expect Substantial ImprovementsNext generation network and Grid system: UltraLightDeliver the critical missing component for future eScience: the integrated, managed network Extend and augment existing grid computing infrastructures (currently focused on CPU/storage) to include the network as an integral component.

  • 10G DataTAG testbed extension to Telecom World 2003 and Abilene/CenicOn September 15, 2003, the DataTAG project was the first transatlantic testbed offering direct 10GigE access using Junipers VPN layer2/10GigE emulation.

    And some statistics about the capacity of networks with Ethernet and the Internet backbone compared.The green line shows specially the increase in available bandwidth with DWDM.It shows at the same time that xxGE slower in time will follow backbone bandwidth.

    Both statistics show clearly that with the old Ethernet technology building frames in the OS is not anymore possible at this level and something has to be done to offload processor or hub internals with clever hardware oriented algorithms. Stabilizing and improving HW/SW drivers for production useE.g. Reducing extra memory copies; interrupt timing moderation