Sponsored by the National Science Foundation Internet2 OpenFlow Backbone Spiral 2 Year-end Project Review Internet2 PI: Eric Boyd Co-PI: Matt Zekauskas Staff: Brian Cashman Students: none directly supported 1-Sep-2010
Sponsored by the National Science Foundation 2 Project Summary The Internet2 OpenFlow Backbone project implements an initial OpenFlow test network (called “Stage0” in our revised proposal) consisting of five small enterprise switches hosted in Internet2 Network PoPs, and connected by wide-area circuits provided by the Internet2 Network and also to enterprise OpenFlow campuses, either directly or through the Internet2 ION service. Review Date: 3-Sep-2010
Sponsored by the National Science Foundation 3 Milestone & QSR Status IDMilestoneStatusOn Time? On Wiki? GPO signoff? S2.aComplete initial deployment planHad a tentative plan in March; worked on NLR interconnect path in Atlanta; now have complete integrated plan for OF, PG, and SN under review by I2 NetSvces >2 IN S2.bLab test OpenFlow 1.0 capable switch and controller Have HP software; now switching to NEC IP8800. Capitalized on IU’s testing. >2 I ?N S2.cDeploy OpenFlow switches to five PoPs Switches ordered, plan under review. Doing a interim Atlanta HP deployment. >2 IN S2.dComplete draft plan for integrating initial deployment with ProtoGENI Plan is to separate backbone networks until PG is comfortable with OF, and aggregate managers interoperate >2 CN S2.eEngineer and test regional network connections Have had discussions with SoX for a direct ATLA connection. ION available. >2 IN S2.fProvide access to OpenFlow slivers for testing Pending>2 I S2.gPlan and engineer GEC9 demoLooking for suggestionsSept. QSR: 4Q2009N/A – prior to contract executionNA QSR: 1Q2010N/A – prior to contract executionNA QSR: 2Q2010Submitted.OTN Review Date: 3-Sep-2010
Sponsored by the National Science Foundation 4 Accomplishments 1: Advancing GENI Spiral 2 Goals The project itself directly relates to three of the GENI Spiral 2 goals: Continuous Experimentation, Integration, and Interoperability. Having a national OpenFlow backbone, run by an organization that provides national research connectivity, available to connect campuses and regional networks, either directly or through the Internet2 ION service, provides a platform for continuous experimentation. In the area of integration and interoperability, it fills in a gap in providing end-to-end OpenFlow services, and a platform to start experimenting with what it means to have multiple organizations cooperate to form a complete end-to-end structure, like multiple Autonomous Systems do for the Internet. In the course of trying to select equipment, the effort has also given companies that build “carrier class” equipment, like Juniper Networks, reasons to start providing OpenFlow as part of their product line. Review Date: 3-Sep-2010
Sponsored by the National Science Foundation 5 Accomplishments 2: Other Project Accomplishments As part of figuring out the OpenFlow plan, we have been working with the GPO to improve connectivity overall for GENI within Internet2 now that the initial Wave MOU with the GPO has expired. We are using some Internet2 OpenFlow project resources to provide the capability for ProtoGENI (and projects connecting through it (Internet-Scale Overlay Hosting (aka Supercharged Planetlab Platform) and Shadownet) to burst above 1Gbps along backbone links. Review Date: 3-Sep-2010
Sponsored by the National Science Foundation 6 Issues Due to the lateness of contract execution (end of May) and the open issues with the end of the initial Internet2 Wave MOU (end of July), we were unable to execute the initial deployment in a timely fashion. Now that a direction has been set, we are prepared to expedite the execution. We are dependent on NEC switch arrival time, as well as ProtoGENI switch arrival time to deploy. We also have limited deployment resources, and OpenFlow, ProtoGENI, and Shadownet deployment touch eight of our nine router nodes. Ideally we would make one trip per node, but may have to make two in some cases to meet other projects’ deadlines. One site that will require at least two visits is Atlanta. We are planning to install a small HP 3500 in Atlanta the second week in September so that the NLR cross-connect can be completed and tested. This switch will also have ION connectivity, so Rutgers and Princeton can start testing of GEC9 demos. This also allows us to test the FlowVisor and Controller. Review Date: 3-Sep-2010
Sponsored by the National Science Foundation 7 Plans What are your plans for the remainder of Spiral 2? –Temporary HP deployment in Atlanta –FlowVisor, SNAC, and Stanford aggregate manager deployed –Testing with Stanford and the GPO site –Deploying the NEC switches –Working with regionals and campuses to ensure required connectivity The GPO is starting to formulate goals for Spiral 3. What are your thoughts regarding potential Spiral 3 work? –Better integration of OpenFlow with the other Internet2-PoP-hosted projects, ProtoGENI in particular. –Explore use of OpenFlow on potential production switches, with a particular focus on supporting slices to allow network researcher testbeds to live on the same infrastructure as the production infrastructure. –Understanding if ProtoGENI can live “on top” of OpenFlow, or vice versa (sharing wide-area bandwidth) Review Date: 3-Sep-2010