ࡱ> '` #bjbj .55555ld77777777}ddddddd$fh@id77777d77d999777}d97}d99R[^77 (58\!_\d0d\j9j<^^8j^\77977777dd9777d7777@ D@  DYNES: A Dynamic Network System for Data Intensive Science Talking Points What: We intend to instrument 40 campuses and 14 Connector sites, each with a low cost Ethernet switch, IDC server (for dynamic circuits), and an FDT server for high throughput (10GE, 1 GE or a few time 1GE), and an attached disk array where needed. Average cost is under $ 20k: $ 20-31k at connectors; $ 11-14k at campuses. We can achieve high throughput with field-proven tools in spite of the low costs, Using the latest switches and servers (based on actual quotes). We will also provide network monitoring via perfSONAR. In-depth monitoring including all the activity in the servers sending and receiving data will be available as well using MonALISA Sites: We are siting the equipment at the 16 US LHC Tier2 and at 24 most active US LHC Tier3 sites (See below). The CMS and ATLAS high energy physics sites also cover heavy ion (nuclear physics) groups. Other fields such as astrophysics (VO, sky surveys) and gravity wave physics (LIGO) also will get direct benefits. At least as important, we will allow researchers at each campus to use the facility to set up dynamic circuits and send/receive data as needed. Overall DYNES is meant to serve the broad data intensive science community in many disciplines. TierCampus [ATLAS]Connector2Boston UniversityNOX2HarvardNOX2IUPUIIndiana Gigapop2Michigan StateMerit2SLACESnet2U ChicagoCIC OmniPoP2U MichiganMerit2U OklahomaGPN2U Texas - ArlingtonLEARN3DukeSOX3Indiana UIndiana Gigapop3TuftsNOX3U California-Santa CruzCENIC3U IllinoisCIC OmniPoP3U Texas DallasLEARN TierCampus (CMS) Connector2CaltechCENIC2MITNOX2PurdueCIC OmniPoP2U California - San DiegoCENIC2U FloridaSOX2U Nebraska - LincolnGPN2U Wisconsin - MadisonCIC OmniPoP3CornellNYSERNet3Florida StateSOX3Florida TechSOX3Johns HopkinsMAGPI3PrincetonMAGPI3RiceLEARN3RutgersMAGPI3Texas TechLEARN3U IowaCIC OmniPoP3U ColoradoUEN3U KansasGPN3U MarylandMAX3U MinnesotaCIC OmniPoP3U Nebraska - OmahaGPN3U TennesseeSOX3UC RiversideCENIC3UCLACENIC3VanderbiltSOX Once the DYNES project (proposed for 2010 2012) is well underway, we will open it up to other campuses that volunteer to bear the (relatively low) cost. Why: Large scale science projects have become distributed using facilities spread over national and intercontinental distances. Researchers need to send/receive datasets of Terabytes (at Tier3s) to tens of Terabytes (at Tier2s) in a reasonable time, which results in data flows lasting hours to days of 1-10 Gbps today, and higher in a few cases. Not only in HEP: several other fields (see for example ESnet network requirements studies) have rising needs that will be of similar scale to HEP within the next 2-5 years. Achieving these flows is no longer a technical barrier: with large numbers of compute nodes and existing production tools at some sites (OSG, gridftp), and at sites with less equipment and expertise using the most recent tools (FDT). But these tools need to be put in researchers hands, at low cost, and the means to use, shared and manage well the available bandwidth must be provided. This is the impetus, and the opportunity to improve and potential transform research, behind DYNES. Requirements to Meet the Needs: Once researchers have this capability, we need A hybrid network to separate support for large flows on circuits, with known (authenticated, authorized) sources and destinations, from the general purpose network that needs to use more expensive carrier-class routers, and firewalls that cannot support many large flows. The ability to allocate guaranteed bandwidth to flows, so that the flows can have predictable throughput, and finish in a predicable time. The ability to fully use the bandwidth allocated (with smooth flows), to prioritize manage the large flows and adjust their characteristics as needed. The research is typically multi-domain, including international. So we need to Establish federated trust (to authenticate and authorize users getting bandwidth over continental and intercontinental distances). Establish a mechanism for negotiation at the inter-domain edges, where each domain can keep control over its resources. Hence no control information sent transparently across a domain (negotiation via a broker is possible. Quantitative Constraints: the Dynamic Circuit Imperative Once there are many sites capable of 1-10 Gbps and in some cases larger flows, nailed up circuits are no longer an alternative: Too many circuits (e.g. a 2 X 2 mesh with 40 sites is a minimum of 780 circuits The aggregate bandwidth is too much, in the 1 to several Terabits/sec range; more in the future There is no way to extend nailed up circuits and hold the bandwidth across multiple domains, especially internationally Overall a nailed up system cannot scale and will immediately become unmanageable. Other advantages of dynamic circuits Mesh protection and reallocation of bandwidth; this is especially flexible and hitless for networks using Layer 1 circuits with active use of VCAT and LCAS One can reconfigure and continue operations in mission-critical networks with limited bandwidth relative to the demands (for example US LHCNet with 4 and soon 6 OC-192 transoceanic circuits). If not for this, then given the outage frequency the level of availability would fall below the specification (due to disruptive outages while the network was reconfigured). Nature of the Project This is a development and integration project: DYNES will build on existing components that are in production use today, including the IDC software suite, applications that switch flows to the circuit-oriented network as needed (such as Terapaths and Lambda station), and perfSONAR for monitoring Higher level services will be built as needed (e.g using MonALISA agent based services which are shown to be robust and run 24 X 7 for months to years in some cases). Benefits Researchers will learn how to handle large datasets using distributed resources R&E networks will learn how to provision, operate and manage networks in support of (many) large flows. Undergoing a paradigm shift (in the circuit part of the hybrid network) following the paradigm shift in science towards large-scale distributed data, processing and collaborative analysis over long distances Benefits It is in this form that Internet2 is likely to realize its potential as a truly advanced network, with an impact that could not previously be foreseen. As such, it could also lead to more general shift towards enabling the at-large community to better deal with increasing quantities of data, and correspondingly increasing data flows. In short, it could have a transformative impact on research and education in the longer term. :;IJKPz ] ^ d  23FGXYuv 01;<QRsthbCJOJQJaJhI/CJOJQJaJhI/CJaJhI/5CJOJQJ\aJhbhI/hI/5hI/ hI/5>* hC5>*hI/hI/5>*B;JK . z { ] ^ f $If $$Ifa$^gdI/gdI/$a$gdI/# qhbb$If $$Ifa$kd$$IflF4 aa6    44 labpytI/ }ww$If $$Ifa$ykd$$IflF''4'6    44 labpytI/ }ww$If $$Ifa$ykd$$IflF''4'6    44 labpytI/  }ww$If $$Ifa$ykd$$IflF''4'6    44 labpytI/  }ww$If $$Ifa$ykd$$IflF''4'6    44 labpytI/&2}ww$If $$Ifa$ykd$$IflF''4'6    44 labpytI/235@F}ww$If $$Ifa$ykdv$$IflF''4'6    44 labpytI/FGITX}ww$If $$Ifa$ykdY$$IflF''4'6    44 labpytI/XY[ou}ww$If $$Ifa$ykd<$$IflF''4'6    44 labpytI/uvx}}ww$If $$Ifa$ykd$$IflF''4'6    44 labpytI/}ww$If $$Ifa$ykd $$IflF''4'6    44 labpytI/}ww$If $$Ifa$ykd $$IflF''4'6    44 labpytI/}ww$If $$Ifa$ykd $$IflF''4'6    44 labpytI/}ww$If $$Ifa$ykd $$IflF''4'6    44 labpytI/}ww$If $$Ifa$ykd $$IflF''4'6    44 labpytI/xrr$If $$Ifa$gdI/ykdq $$IflF''4'6    44 labpytI/ "*0qhbb$If $$Ifa$kdT$$IflF4 aa6    44 labpytI/0137;}ww$If $$Ifa$ykd[$$IflF''4'6    44 labpytI/;<>EQ}ww$If $$Ifa$ykd>$$IflF''4'6    44 labpytI/QRTms}ww$If $$Ifa$ykd!$$IflF''4'6    44 labpytI/stv}ww$If $$Ifa$ykd$$IflF''4'6    44 labpytI/}ww$If $$Ifa$ykd$$IflF''4'6    44 labpytI/}ww$If $$Ifa$ykd$$IflF''4'6    44 labpytI/}ww$If $$Ifa$ykd$$IflF''4'6    44 labpytI/}ww$If $$Ifa$ykd$$IflF''4'6    44 labpytI/}ww$If $$Ifa$ykds$$IflF''4'6    44 labpytI/,-:;KL_`vw ()J^$=WpKXm 8"@"A"###ĿĿ׸׸ĸװ hbhn~ vVWb:N & Fgdb8^8gdb & Fgdb & Fgdb & Fgdb & Fgdb & Fgdbgdb !8"A"B""##gdI/ & F gdEQRTmstv&,-/4:;=EKLNY_`bjvwy      $ ( ) * H J =>n~ vVWb:N8AB000000000000000 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0000 00 000000 0 0 0 0 0 0 0000 0 0 0 0 00 0 0 00 0 0 00 00 0 0 00 0 00  &235@FGITXY[ouvx} "*0137;<>EQRTmstv&,-/4:;=EKLNY_`bjvwy      $ ( ) A00m00m00m@0 00m00m00m@0 00m00m00m@0 00m00m00m@0 00m00m00m@0 0 0 m0 0 m0 0 m@0 0 0 m0 0 m0 0 m@0 00 m00 m00 m@0 00 m00 m00 m@0 00 m00 m00 m@0 00m00m00m@0 00m00m00m@0 00m00m00m@0 00m00m00m@0 00m00m00m@0 00 00 00 @0 00L00L00L@0 00L00L00L@0 00L00L00L@0 00L00L00L@0 00L00L00L@0 0 0L0 0L0 0L@0 0 0L0 0L0 0L@0 00L00L00L@0 00L00L00L@0 00L00L00L@0 00L00L00L@0 00L00L00L@0 00L00L00L@0 00L00L00L@0 00L00L00L@0 00 L00 L00 L@0 0 0 L0 0 L0 0 L@0 0"0 L0"0 L0"0 L@0 0$0 L0$0 L0$0 L@0 0&0 L0&0 L0&0 L@0 0(0L0(0L0(0L@0 0*0L0*0L0*0L@0 0,0L0,0L0,0L@0 0.0L0.0L0.0L@0 000L000L000L@0 020 020 020 @0 @ 0 000#/ 2FXu0;Qs,:K_v(=# !"#$%&'()*+,-.0123456789:;<=>?@A#8@0(  B S  ?T\nt*1IPnu  \ a x  ,/6>3;K) H J W8B hOZ1[5cF>G!^Q`+R ABT%FMH)-TCo^_";C%tgؐ,m!lR%|~vbh^`OJQJo(hHh ^`o(hH.hpp^p`OJQJo(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJQJo(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJQJo(hHhhh^h`OJQJo(hHh88^8`OJQJ^Jo(hHoh^`OJQJo(hHh  ^ `OJQJo(hHh  ^ `OJQJ^Jo(hHohxx^x`OJQJo(hHhHH^H`OJQJo(hHh^`OJQJ^Jo(hHoh^`OJQJo(hHh ^`o(hH.h ^`o(hH.hpp^p`OJQJo(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJQJo(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJQJo(hHh^`OJQJo(hHh ^`o(hH.hpp^p`OJQJo(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJQJo(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJQJo(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohpp^p`OJQJo(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJQJo(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJQJo(hHh  ^ `OJQJo(hHhxx^x`OJQJo(hHhHH^H`OJQJo(hHh^`OJQJo(hHh^`OJQJ^Jo(hHoh^`OJQJo(hHh^`OJQJo(hHhX X ^X `OJQJ^Jo(hHoh(#(#^(#`OJQJo(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohpp^p`OJQJo(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJQJo(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJQJo(hHh ^`o(hH.h ^`o(hH.hpp^p`OJQJo(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJQJo(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJQJo(hHh88^8`OJQJo(hHh^`OJQJ^Jo(hHoh  ^ `OJQJo(hHh  ^ `OJQJo(hHhxx^x`OJQJ^Jo(hHohHH^H`OJQJo(hHh^`OJQJo(hHh^`OJQJ^Jo(hHoh^`OJQJo(hHh^`OJQJo(hHh ^`o(hH.hpp^p`OJQJo(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJQJo(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJQJo(hHh 88^8`o(hH.h^`OJQJ^Jo(hHoh  ^ `OJQJo(hHh  ^ `OJQJo(hHhxx^x`OJQJ^Jo(hHohHH^H`OJQJo(hHh^`OJQJo(hHh^`OJQJ^Jo(hHoh^`OJQJo(hHh 88^8`o(hH.h^`OJQJo(hHh  ^ `OJQJo(hHh  ^ `OJQJo(hHhxx^x`OJQJ^Jo(hHohHH^H`OJQJo(hHh^`OJQJo(hHh^`OJQJ^Jo(hHoh^`OJQJo(hH [+%F)-TG!C%tgc%|~ Ao^_hO,m!l                                                                                                             esRTFZkIY =EIU-X8`EQRTmstv&,-/4:;=EKLNY_`bjvwy      $ ( ) @ @Unknowngz Times New RomanTimes New Roman5Symbol3& z Arial7&{ @CalibriO5  hMS Mincho l r   ;Wingdings?5 z Courier New"qhFF+2+2Y242QHX ?I/2:DYNES: A Dynamic Network System for Data Intensive Science Harvey Newman Harvey Newman<         Oh+'0 , DP p | <DYNES: A Dynamic Network System for Data Intensive ScienceHarvey Newman Normal.dotHarvey Newman2Microsoft Office Word@@na@na+՜.+,0( hp  Caltech2' ;DYNES: A Dynamic Network System for Data Intensive Science Title  !"#$%&'()*+,-./0123456789:;<=>?@ABDEFGHIJKLMNOPQRSTUWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~Root Entry F*Data C%1TableVNjWordDocument.SummaryInformation(DocumentSummaryInformation8CompObjq  FMicrosoft Office Word Document MSWordDocWord.Document.89q