GENI Spiral 1 at six months Taking stock

GENI Spiral 1 at six months Taking stock, looking ahead GENI Engineering Conference 4 Miami, Florida Chip Elliott April 1, 2009 www. geni. net Sponsored by the National Science Foundation

Outline • Taking Stock of GENI’s Current Status – – Net. SE Research Agenda GENI Spiral 1 Ongoing “Vertical” and “Horizontal” Integrations GENI Solicitation 2 • GENI Spiral 2 – Let’s start the discussions – Vertical – Simple ‘clearinghouse’ functionality – Horizontal – Building out pervasive, on-demand Layer 2 connectivity • The next two days Sponsored by the National Science Foundation April 1, 2009 2

GENI supports Fundamental Challenges in Network Science & Engineering (Net. SE) Science Understand the complexity of large-scale networks - Understand emergent behaviors, local–global interactions, system failures and/or degradations - Develop models that accurately predict and control network behaviors Technology Develop new architectures, exploiting new substrates - Develop architectures for self-evolving, robust, manageable future networks - Develop design principles for seamless mobility support - Leverage optical and wireless substrates for reliability and performance - Understand the fundamental potential and limitations of technology Society Network science and engineering researchers Distributed systems and substrate researchers Enable new applications and new economies, while ensuring security and privacy Security, - Design secure, survivable, persistent systems, especially when under attack - Understand technical, economic and legal design trade-offs, enable privacy protection - Explore AI-inspired and game-theoretic paradigms for resource and performance optimization Sponsored by the National Science Foundation April 1, 2009 privacy, economics, AI, social science researchers 3

Net. SE Research Agenda GENI Science Plan Why We Dream of GENI Sponsored by the National Science Foundation April 1, 2009 4

Kudos to the Net. SE Council Ellen Zegura (Chair) Tom Anderson (UW) Joe Berthold (Ciena) Charlie Catlett (Argonne) Joan Feigenbarum (Yale) Stephanie Forrest (UNM) Mike Dahlin (UT Austin) Jim Hendler (RPI) Michael Kearns (U. Penn) Ed Lazowska (UW) Chip Elliott (GPO) Peter Lee (CMU) And not shown. . . Roscoe Giles Helen Nissenbaum Larry Peterson (Princeton) Sponsored by the National Science Foundation Jennifer Rexford (Princeton) April 1, 2009 Alfred Spector (Google) 5

Capturing discussion and design GENI System Engineering Documents Sponsored by the National Science Foundation April 1, 2009 6

GENI Spiral 1 • We are here Multiple competing control frameworks, beginnings of an at-scale experiment plane GENI Prototyping Plan Sponsored by the National Science Foundation April 1, 2009 7

Spiral 1 “Vertical” and “Horizontal” Integrations Create my slice GENI Goal #1 Clearinghouse Fund multiple, competing teams to develop GENI Clearinghouse technology, encourage strong competition within the first few spirals Goal #2 Demonstrate end-to-end slices across representative samples of the major substrates / technologies envisioned in GENI Components Aggregate A Aggregate B Backbone Net “Horizontal Integration” Aggregate C Computer Cluster “Vertical Integration” Metro Wireless Sponsored by the National Science Foundation April 1, 2009 8

Key goals for Spiral 1 Clusters • Integrate “vertically” into your control framework • Integrate “horizontally” to create end-to-end slices Reference Design Components Aggregate A Aggregate B Aggregate C Computer Cluster Backbone Net Metro Wireless • Demonstrate (early) integrated prototypes up and stumbling within the next 3 -6 months • . . . and design GENI as you go! Sponsored by the National Science Foundation April 1, 2009 9

General feedback on GENI prototyping thus far • GENI prototyping is getting strongly positive reviews from. . . – – Wise elders Research community organizations Campus and national infrastructure communities Networking and computer industry • General tenor: GENI prototyping teams are rocking! – – – Many of the best systems researchers in the US are actively engaged People are working together towards very ambitious, leading-edge goals Rapid prototyping with academic/industrial teams is exactly right A level of excitement we haven’t seen in decades “It’s like DARPA in the good old days” • But of course we have some issues to address. . . – – – Cross-project integration is (always) hard 2 -3 projects seem to be making little or no progress Working groups may not, in fact, be working Experimentation wishlist is still murky Thursday morning we’ll discuss with open microphone Sponsored by the National Science Foundation April 1, 2009 10

GENI Solicitation 2 – Many topnotch proposals How are we doing? Security Experiments Total: 91 proposals received Sponsored by the National Science Foundation April 1, 2009 11

Looking forward to GENI Spiral 2 • What will be in Spiral 2? • Let’s begin to discuss – No decisions yet! – Just starting discussions • GPO wishlist – – – Measurements Simple clearinghouse Federation (early try) Layer 2 buildout Early experiments GENI Prototyping Plan Sponsored by the National Science Foundation April 1, 2009 12

Instrumentation and Measurement • “GENI will be well-instrumented. . . ” – – – Scope is still quite unclear Few funded efforts to date, few new proposals What kinds of instrumentation? Archival? Analysis? Sharing? Opt-in users bring privacy issues • Needs vigorous investigation / prototyping – Paul Barford chairing GPO-sponsored workshop – GPO will be pushing more efforts in this area – Your thoughts and ideas are needed here! Sponsored by the National Science Foundation April 1, 2009 13

Thoughts on simple clearinghouse functionality Spurred by recent discussions on the working group list Standard issue BBN napkin Sponsored by the National Science Foundation April 1, 2009 14

Major functions of a clearinghouse, as conceived through an ‘e-commerce’ analogy 1. Bring ‘buyers’ and ‘sellers’ together. 2. Keep accurate records of all ‘purchase’ transactions, so we know who controls what resources at which times. Transaction log “I’ll buy it” Identified Aggregates Researchers Reduce N x M trust relationships to N+M Enable forensics and big red “kill the experiment” button Sponsored by the National Science Foundation April 1, 2009 3. Implement systemwide policy as needed. No, you can’t have all resources “I want them all!” Identified Aggregates Enable system-wide usage policies as the need arises. 15

What does a clearinghouse NOT do ? 1. Understand, describe, or manipulate resource descriptions (rspecs). 2. Provide experiment workflow tools. 3. Participate in the “life of an experiment. ” In short. . . The clearinghouse is a transaction-processing system. . . Much like Pay. Pal. . . It gets invoked after the researcher has decided what to buy (e, g after the research has put resources in a ‘shopping cart’. . . And closes the ‘purchase’ transaction That’s all. Sponsored by the National Science Foundation April 1, 2009 16

Establishing trust and identity relationships 1. Bring ‘buyers’ and ‘sellers’ together. Identified Aggregates Researchers Reduce N x M trust relationships to N+M Sponsored by the National Science Foundation Sketch of an approach. . . 1. Aggregates ‘affiliate’ with clearinghouse, via out of band channel involving people. 2. Clearinghouse forms authenticated channel to aggregate (eg via SSL); this channel will be used for all their subsequent communication. 3. Researchers ‘log in’ to clearinghouse via web page interface hosted at clearinghouse. 4. Clearinghouse uses Shibboleth to identify the researcher. 5. While researcher is logged in, clearinghouse uses authenticated channel to researcher’s web browser for all communication with that researcher (eg via SSL). April 1, 2009 17

Transaction processing 2. Keep accurate records of all ‘purchase’ transactions, so we know who controls what resources at which times. Transaction log tickets “I’ll buy it” Enable forensics and big red “kill the experiment” button Sponsored by the National Science Foundation 1. Researcher accumulates desired resources in a ‘shopping cart’ for a given slice [this idea still needs much work] 2. Researcher clicks equivalent of ‘purchase’ button. 3. Clearinghouse runs system-wide policy checks as needed, may reject the purchase. 4. Clearinghouse makes ‘purchase’ requests from aggregates to obtain all resources still available. 5. Clearinghouse makes high-reliability log of transaction, and also maintains current set of resources allocated to that slice. 6. Clearinghouse delivers purchased resources to researcher by means of ‘tickets’ for their later use. 7. Clearinghouse gives matching tickets to resources’ aggregate / component managers. April 1, 2009 18

Enforcing system-wide policy as needed 3. Implement systemwide policy as needed. No, you can’t have all resources “I want them all!” Sketch of an approach. . . 1. Researcher puts resources in slice shopping cart, clicks ‘purchase’. 2. Clearinghouse knows identity and attributes of researcher. 3. Clearinghouse also knows identity of each aggregate. 4. “Somehow” clearinghouse policy engine does basic sanity checks. 5. Since I don’t understand this area, clearly I should not be describing how this all works! Identified Aggregates Enable system-wide usage policies as the need arises. Sponsored by the National Science Foundation April 1, 2009 19

Manipulating slices. . . Researchers in this slice: PI: Person A Can modify membership: Persons B, C Can add/delete resources: Persons D, E Can use slice: Persons F, G, H, I Resources in this slice: Aggregate X, resources x 1, x 2, … Aggregate Y, resources y 1, y 2, . . . Shopping cart for this slice: Aggregate X, resources x 1, x 2, … Aggregate Y, resources y 1, y 2, . . . Slice records (a variant version) “I’ll buy it” Most control frameworks do this well. Sponsored by the National Science Foundation April 1, 2009 20

And one more major function: Federation It’s likely to be tough, so let’s start in Spiral 2 NSF GENI Sponsored by the National Science Foundation April 1, 2009 21

Building out pervasive, on-demand Layer 2 connectivity Sponsored by the National Science Foundation April 1, 2009 22

Currently in the works Prototyping GENI through campuses • August Meeting at O’Hare – Thanks to Edu. Cause (Mark Luker, Garret Sern) – Stimulated by Larry Landweber • CIOs from 11 major research universities – Berkeley, Clemson, GA Tech, Indiana, MIT, Penn State, Rice, U. Alaska, UIUC, UT Austin, U. Wisconsin • Discussions of representative GENI prototypes – Nick Mc. Keown, Stanford (Open. Flow) – Arvind Krishnamurthy, UW (Million Node GENI) – GPO Staff • Near-term GENI / CIO activities – How to “GENI-enable” campus IT infrastructure – Coordinated policy for handling side-effects of network research (Larry Peterson, Helen Nissenbaum) Sponsored by the National Science Foundation April 1, 2009 23

The Next Two Days • Tackle design issues – In the Working Groups – Locus of all GENI design activities (arguments) – Deliberately cut across “research silos” • Tackle integration issues – In the control framework “clusters” – Vertical & horizontal integration – Source of insights and raw experience • GPO System Engineers in the mix Sponsored by the National Science Foundation April 1, 2009 24

Some thoughts as we begin. . . • Move fast – A new world is unfolding very quickly • Think big – We have enormous opportunities • Work together – GENI prototyping is very much a positive-sum game • Be yourself – We value your creativity and insights Sponsored by the National Science Foundation April 1, 2009 25