NSF LC Workshop Comments

Transcription

1 NSF LC Workshop Comments John Wroclawski USC/ISI

2 The DETER Testbed Facility is: An experimental research platform... that spports the design, constrction, exection and stdy of complex networked cyber- and cyberphysical system scenarios in ways that the real world often does not. Easily designed, implemented, and managed, centrally, by a single person or team. Open, modifiable, extensible, measrable, and (re)configrable. With clear visibility into both aggregate and individalcomponent behavior. When appropriate, repeatable and predictable. 2

3 The DETER Project is significantly more: A program of research in experimental cybersecrity research infrastrctre research on research Development and operation of the DETER Testbed Facility (DETERLab) A program of commnity otreach, evangelization, and edcation DETERLab itself is operated as A national and international resorce for cybersecrity experimentation A technology transfer vehicle for the research program 3

4 Random Cool Pictre 100K-Node Worm/Botnet/DDOS Scenario (2012) (hh, looks familiar..) 4

5 DETER Technical Synopsis Core Emlation/Simlation/Modeling Engine Based on Emlab for resorce management, allocation, sharing ~ 700 nodes, > 5000 cores, two sites (USC/ISI and UC Berkeley) Mlti-resoltion modeling spport (variable-fidelity virtalization) Heterogeneos resorce/testbed federation architectre Scalable experiment orchestration and control Mlti party experiment spport (realistic limited visibility) Hman behavior modeling sbsystem Network traffic modeling, captre, and generation facilities Data collection and visalization spport For a concise frther overview, see DETER smmary chapter in The GENI Book

6 Scale, Reach and Impact DETER Research Commnity: Over 750 researchers From ~180 organizations in 35 contries ~200 pblished papers listed at DETER Partner Facilities: Pacific Northwest National Labs (PNNL) Defense Research Department Canada (DRDC) Israel National Cyber Brea Washington State University Singapore National University Vanderbilt University DETER Edcational Commnity ~ 9000 niqe stdent sers Representing ~140 distinct classes (many offered repeatedly) at 91 niqe instittions: ª Major research niversi/es ª commnity colleges and minority-serving ins/t/ons ª A few high schools! ª Approximately 20% interna/onal

7 Observation 1 Scope of the Problem

8 Large Scale Experiments - The State of the Art or A brief history of cyber infrastrctre in the large scale / networked system / etc. domain, with generalizations Define a meaningfl experiment Represent yor experiment Rn yor experiment -> Big Problem! (Room for Advance) -> Some effort Good Progress! -> Mch effort Excellent Progress!

9 Two Distinct Types (of Experiment) Deployment / Exploratory Stdy ( Open Ended ) Rigoros Hypothesis Test Althogh there s obvios overlap (and some folks are aiming to do both at once), these two objectives lead to noticeably different reqirements, and shold not be confsed.

10 Open Ended Stdy Early-stage or prototype experimental deployment In manageable conditions For prpose of better nderstanding fitness for prpose, basic design decisions, sability, effective limits, etc. Key isses Appropriate Realism Anticipating Evoltion

11 Rigoros Hypothesis Test The big problem here is identifying the worst case Remember rare events Second problem: the meaning of repeatability Cannot be 100% concrete, bit-level repeatability. Mst be driven by abstraction. Significant intellectal challenge To motivate: need framing problems where competent worstcase and repeatability analysis is: In reach and illstrative Can be reasoned abot and captred somewhat rigorosly, and sed to drive strategic infrastrctre research and development

12 Observation 2 Centrality of Architectre

13 Architectre high level design principles that gide the technical development of a system, especially the engineering of its protocols and algorithms Two distinct levels: A specification of system modlarity, fnctional decomposition, interface locations and definitions, etc. the what. A set of fndamental strctring principles that drive the choices of the first bllet the why.

14 Architectre Fnctions Basic system technical reqirements and properties Distinction between reqirements / properties of the architectre and reqirements / properties of a system. Many systems can share one architectre Coordination in space mltiple independent implementers / contribtors Coordination over time mltiple, evolving technologies and ses

15 Pt another way.. Crcial, minimal fixed points that deconstrain the rest of the system s design and implementation For increased flexibility, capability, seflness, & breadth of participation Long lived relative to the rest of the system design To increase overall lifetime, speed potential evoltion, facilitate move to new capabilities and technologies, etc

16 So where does all this wonderflness come from? Several possibilities, from pre a priori to gided qasi-emergent. Bt! Not by accident! Not by happenstance.. Not by comptational archaeology..

17 Observation 3 Fnding and Sstainability

18 DETER Project Fnding Model Explicit two-level Core / Satellite fnding strategy Core fnding spport from NSF (1 st 4 years) and DHS (ongoing) Stable base spport for all three project activities Satellite fnding throgh direct project participation in other fnded research programs Primarily DARPA, DHS, some NSF Project personnel take on expert, shepherd and/or playing field designer roles in others projects Sbstantial financial and intellectal benefit back to the DETER project

19 Fnding and Sstainability A two level challenge First: Evoltionary sstainability As one example DETER s fnding strategy has provided strong spport for ~12 years, while continally advancing both facility capabilities and state of the art Second: Generational advance At some point in any infrastrctre s lifetime, generational change will likely become necessary or appropriate Think clearly abot generational vs. evoltionary bondaries, circmstances and timeframes. Architectre helps J Recognize, evangelize, and plan ahead.

20