1. Linda Winkler IWS2000/APAN Technology Session Feb 17, 2000 linda@startap.net STARTAP QBONE Activities

3. Internet2 QBone Initiative Build interdomain testbed infrastructure Experiment and improve understanding of DiffServ Incrementally improve testbed Support intradomain & interdomain deployment Lead and follow IETF standards work Some parts of DiffServ architecture ready; others far from it Our experience will inform standards process Openness of R&E community gives us an edge Users will tolerate the flakiness of an experimental infrastructure Engineers will share experience and measurement data QBone as infrastructure that only a graduate student could love

4. QBone Architecture (10km view) IETF Diff (EF PHB) + QBone Serv (QPS) QBone Premium Service Idea: converge on Jacobsons VLL Premium service Well-defined SLS: Peak rate (R) & Service MTU (M) implying a token bucket meter Near-zero loss Low jitter Delay variation due to queuing effects should be no greater than the packet transmission time of a M sized packet All bets are off if the reserved interdomain route flaps Plus important value-adds: Integrated measurement/dissemination infrastructure Experimentation with pre-standards inter-domain bandwidth brokering and signaling

6. STARTAP Architecture Diagram Ethernet Various SpeedsOC3c Ameritech NAP Ascend 5XX ATM switches U.S. and International Network Connections oooooo 6TAP Router (IPv6) Cisco 7200 STAR TAP Router (IPv4) Cisco 7505 DS Cisco 7507 NLANR AMP NLANR Web Cache Star Tap General Purpose UNIX Host NLANR OC3MON TransPAC STAR TAP Cisco LS1010 ATM switch

7. STARTAP Simplified AADS STARTAP IPV4 IPV6 IPV4 DS JP SingAREN Abilene ESnet IPV4 DS IPV4 DS Argonne IPV4 DS APAN KR 1 Mbps Mar 00

8. STARTAP- IP DiffSrv Router One stop shopping for STARTAP DS routes STARTAP DS Router is AUP free STARTAP DS Router will run experimental Cisco IOS Setting up Bandwidth Broker/Resource Manager

9. DS Components Expedited Forwarding Cisco model Ingress (classify, mark, police) CAR Egress (bandwidth allocation) WFQ Committed Access Rate Traffic Conditioner Meter, mark and police (drop) Queueing WFQ PQ GARA Bandwidth Broker Admission control Resource management Router configuration Traffic Shaping (GTS, DTS)

10. DOE EMERGE Testbed www.evl.uic.edu/cavern/EMERGE

11. Globus Grid Services Package Deploy standard infrastructure across participating sites Provide maximum capabilities to applications Increase what can be taken for granted when developing applications Reduce deployment burden at sites Services include Authentication Resource discovery Resource management ( Globus Architecture for Reservation and Allocation – GARA) Instrumentation (netlogger)

12. Globus Architecture for Reservation and Allocation (GARA) Three important contributions: Support for Advance Reservations Uniform treatment of underlying resources Network, disk, CPU etc… Currently supported: Differentiated Services, DPSS, and DSRT Layered architecture enables generic co- reservation agents User is able to select API best suitable for his need

13. GARA BasicsReservations There is a generic reservation, which has several properties: Start Time (now or future) and Duration Resource type/Underlying resource identifier Resource-specific (bandwidth, % CPU…) All reservations are treated uniformly: Create/Modify (Given properties) => Returns Reservation Handle Destroy Monitor (Callbacks or Polling)

14. End-to-End Network Reservations Algorithm for a single network reservation Discover which gatekeepers/resource managers need to be contacted (MDS) Make reservation at each one. User process Workstation Resource Manager MDSEnd-to-End Co-Reservation Agent

15. GARA features Reservation-Subtypes low-latency Used to support jitter sensitive applications using the same EF aggregate behavior as high-throughput TCP applications Based on traffic shaping and Priority-Queuing Currently being evaluated background Used to support bulk transfers, including deadline support. Takes at least the amount of premium traffic required to fulfill the deadline; more if not used actively Monitoring of the network edges Provides feedback to applications when they send too fast

16. Basic Experiment I Goal: Proof of Premium Quality for TCP flows Run modified ttcp with different traffic and reservation profiles; gather receiver bound statistics Create congestion using gen_send/recv

17. Basic Experiment II Use tcpdump and tcptrace understand the behavior Use netlogger_enabled ttcp_gara and gen_recv to publish data into MDS. Real Time Visualization of rtt

18. Basic Experiment III Goal: Demonstrate Premium Quality for UDP flows Run mgen/drec with different traffic and reservation profiles; gather receiver bound statistics Create congestion using gen_send/recv

19. Basic Experiment IV Goal: Proof Premium Quality for UDP flows sharing one EF aggregate Run mgen/drec with different traffic and reservation profiles; gather receiver bound statistics Run high-throughput ttcp using a correct reservation Create congestion using gen_send/recv

20. Testing Perform local ANL/GARNET tests Each EMERGE site provides a test machine (Linux RH6.1) Phase 1- measure baseline end-to-end BE performance (desire 100 Mbps ~ achieving 70-80 Mbps) Phase 2- measure single stream end-to-end Premium service Need better measurement tools and techniques Phase 3- measure application end-to-end performance Phase 4- measure multiple stream end-to-end Premium service Phase 5- evaluate a MPLS label-switched router as a Diffserv router

21. References www.globus.org www.globus.org/qos www.internet2.edu/qos