Data TAG overview 23 January 2003 SWITCH-cc meeting

Data. TAG overview 23 January 2003 SWITCH-cc meeting 1

Summary u Why Data. TAG? u Data. TAG project u General information u Test-bed extensions u Network map u Research topics u Issues u Acknowledgements 23 January 2003 SWITCH-cc meeting 2

Background for Data. TAG u u u CERN is building the LHC (start-up in 2007? ): an unprecedented amount of data will have to be analyzed and CERN alone will not have enough computing resources to process it The planned computing model is distributed geographically (GRID) The EU-Data. Grid project is addressing the middleware problem Reliable, advanced networking is needed underneath Traffic will not look like any IP commodity traffic now A lot of bandwidth is necessary, but it won’t be enough 23 January 2003 SWITCH-cc meeting 3

Addressing the problem u Buy a lot of bandwidth u Buy (expensive) network equipment u Address security issues without compromising performance (firewalls) u Tune some more or less obvious TCP parameter u Design the network for reliability and performance u Plan for interconnection with key research networks 23 January 2003 SWITCH-cc meeting 4

Why network research? u u u Need to enhance data transport protocols (TCP) Need to measure the perceived application performance Need to try the GRID software in a WAN environment and at high speed: what is actually going wrong if things go wrong in the network? Need to test the compatibility between EU and US GRIDs (middleware integration) Need to test network technology with enough speed and enough features to support the planned GRID workload (end-to-end inter-domain Qo. S) 23 January 2003 SWITCH-cc meeting 5

Data. TAG project u u u “Research and technological development for a transatlantic GRID” (project full title) IST project (EU funded), supported by the NSF and the Do. E (Caltech) Partners: PPARC (UK), INRIA (FR), University of Amsterdam (NL), INFN (IT) and CERN (CH) Researchers from Caltech, SLAC and Canada Test-bed kernel: a transatlantic STM-16 (Tsystems) between Geneva (CERN) and Chicago (Star. Light), with interconnected workstations 23 January 2003 SWITCH-cc meeting 6

General information u 2 -years project: 2002 and 2003 u Dedicated staff (recruited on project budget) u Part-time staff, shared with other activities u u Open to cooperate with other projects: EUData. Grid, GEANT, Abilene, Tera. Grid, Nether. Light, etc. Typical EU work package structure: quarterly reports, deliverables at fixed deadlines, periodic reviews by external inspectors. Very formal and structured framework, but effective to avoid project drifting, delays and wastes 23 January 2003 SWITCH-cc meeting 7

Data. TAG Network map R 06 chi-Alcatel 7770 W 01 chi w 02 chi w 03 chi w 04 chi w 05 chi w 06 chi ONS 15454 SURFNET 2 x 1 GE CANARIE V 10 chi v 11 chi v 12 chi v 13 chi 4 x 1 GE 8 x 1 GE Alcatel 1670 R 05 chi-Juniper. M 10 SUNNYVALE ABILENE Stm 64(L 3) 10 GE 2 x 1 GE Alcatel 1670 2 x 1 GE VTHD/INRIA 1 GE Stm 16 (France. Telecom) 1 GE 2 x 1 GE Stm 16(DTag) R 04 chi-Cisco 7609 R 04 gva-Cisco 7606 1 GE R 05 gva-Juniper. M 10 Extreme Summit 1 i 1 GE Teragrid Juniper. T 640 2 x 1 GE Cisco 5505 -management Chicago Geneva 1 GE Cisco 2950 -management ar 3 -chicago -Cisco 7606 Stm 4(DTag) Cernh 4 -Cisco 7609 Data. TAG 1 GE CERN External Network 1 GE Vlan 4 Vlan 5 Vlan 7 Stm 16(GC) 4 x 1 GE 2 x 1 GE Extreme Summit 5 i 10 GE SURFNET W 03 gva w 04 gva 1 GE 10 x 1 GE ONS 15454 W 01 gva w 02 gva w 05 gva w 06 gva ONS 15454 2 x 1 GE R 06 gva-Alcatel 7770 SWITCH 1 GE Stm 16(Swisscom) Cernh 7 -Cisco 7609 GEANT Stm 16(Colt) backup+projects GARR/CNAF edoardo. martelli@cern. ch - last update: 20021204

Data. TAG Routing map W 01 chi w 02 chi w 03 chi w 04 chi w 05 chi w 06 chi SURFNET Management addresses and path to reach them from the internet Datatag testbed addresses Path of the CCC tunnel from CNAF Path between the OC farms in Chicago and Geneva Path between Sunnyvale and the PC farm in Geneva R 06 gva R 06 chi V 10 chi v 11 chi v 12 chi v 13 chi W 01 gva w 02 gva w 05 gva w 06 gva SURFNET W 03 gva w 04 gva CANARIE VTHD/INRIA R 05 gva R 05 chi SUNNYVALE ABILENE R 04 gva R 04 chi 192. 91. 236. 0/23 192. 91. 238. 0/23 Cisco 5505 -management Teragrid Juniper. T 640 Chicago Geneva 192. 91. 246. 192/26 192. 91. 244. 0/27 Data. TAG CERN External Network SWITCH Cisco 2950 -management edoardo. martelli@cern. ch - last update: 20021204 ar 3 -chicago Cernh 4 Cernh 7 GEANT GARR/CNAF

(some) Research topics (I) u u u Linux kernel tuning for high performance: for example, 8 Terabytes in 24 hours, memory-tomemory (achieved by S. Ravot, Caltech) Bulk file transfer (Terabyte disk-to-disk, at 2 Gbps, achieved by Canadian researchers between TRIUMF and CERN) TCP stack improvements (things get worse with longer RTT) Application-level performance measurement GRID middleware interoperability between EU and US 23 January 2003 SWITCH-cc meeting 10

(some) Research topics (II) u u 10 Gb tests (10 Gb. E and transatlantic link upgrade, tentatively in September 2003) at layer 2 and layer 3 Multi-vendor equipment (Alcatel 1670 and 7770, Cisco 760 x, Juniper M 10, Extreme Summit switches) u Qo. S tests, advanced reservation u Optical networking: validation of equipment 23 January 2003 SWITCH-cc meeting 11

Issues (I) u u Electrical power in Chicago (!!!) Broken network hardware (mainly 10 Gb. E cards, but not only) u Broken PCs u Hung routers u Never enough workstations available for testing u u u Reservation system: it will never be sophisticated enough Network topology: each research group has different requirements Alcatel 1670: needs STM-16 reconfiguration 23 January 2003 SWITCH-cc meeting 12

Issues (II) u u Cisco 760 x components (un)availability Demo workshops (i. GRID 2002, SC 2002, …): nice to see, but organizational nightmares KPNQwest collapse (re-procurement via T-systems) Routing: interconnections with CERN production network, with external production networks, with external test networks, even with commodity Internet; furthermore, routing is open for experiments u Management access to devices u OOB access (Chicago) 23 January 2003 SWITCH-cc meeting 13

Issues (III) u u u A lot of enthusiasm and interest, not always supported by adequate planning When adequate planning is there, it is ignored (because of the enthusiasm, of course) Partners coordination is not straightforward: mix of shared and private resources DNS, security, access control, management servers, VLANs, WWW site, IP addressing (>100 addresses assigned), etc. : all need planning, work and maintenance Po. P management (run out of rack space) + installation issues (mainly, but not only, Alcatel) 23 January 2003 SWITCH-cc meeting 14

Acknowledgements u u u Data. TAG is for what you would have always liked to do and you never dared to do because it was too risky, or too expensive, or too complicated, or because you could not afford a test laboratory Many thanks to DANTE/GEANT and Do. E/Caltech, who are actively supporting the project http: //www. datatag. org 23 January 2003 SWITCH-cc meeting 15