tools for the analysis and design of complex multi scale
play

Tools for the Analysis and Design of Complex Multi-Scale Networks: - PowerPoint PPT Presentation

Tools for the Analysis and Design of Complex Multi-Scale Networks: Overview MURI Annual Review Berkeley, October 27, 2011 J. Walrand, PI SLIDES: http://web.me.com/jeanwalrand/Site_3 1. Team Venkat


  1. Tools for 
 the Analysis and Design of Complex Multi-Scale Networks: 
 Overview � • MURI Annual Review 
 Berkeley, October 27, 2011 � • J. Walrand, PI � SLIDES: http://web.me.com/jeanwalrand/Site_3

  2. 1. Team � • Venkat Anantharam, UCB • John Doyle, Caltech • Weibo Gong, UMASS • Steven Low, Caltech • R. Srikant, UIUC • Don Towsley, UMASS • Jean Walrand, UCB • Avideh Zakhor, UCB • 14 PhD Students, 7 Postdocs • 12 Journal Papers, 21 Conference papers

  3. 2. Scientific Objectives and Technical Approach �

  4. 2. Scientific Objectives and Technical Approach � High Throughput Low Delay Security & Reliability GOALS of PROPOSED WORK: Improve Understanding of Multi-Scale Features of Networks Use that Understanding to Develop Better Protocols

  5. 2. Scientific Objectives and Technical Approach � High Throughput Low Delay Methodology for LRD and Performance Evaluation Protocols Security Game: Attack/Defense Reliability

  6. 3. Scientific Issues � Local Information - Global Impact Multiple Time Scales Uncertainty Security & Reliability

  7. 4. Scientific Accomplishments � High Throughput Low Delay Methodology for LRD and Performance Evaluation Protocols Security Game: Attack/Defense Reliability

  8. 4. Scientific Accomplishments � High Throughput Low Delay Network Sharing Bound Power Laws in Network Analysis 
 Methodology for LRD and Performance Evaluation Fat tails, hard limits, thin layers Protocols Security Game: Attack/Defense Reliability

  9. 4. Scientific Accomplishments � High Throughput Low Delay Network Sharing Bound Power Laws in Network Analysis 
 Methodology for LRD and Performance Evaluation Fat tails, hard limits, thin layers Access Delay in IEEE 802.11 Stability of a P2P Protocol Protocols Target Tracking with In-Network Aggregation 
 Connection-Level Scheduling 
 E fg ect of Bu fg ers on Stability Security Game: Attack/Defense Reliability

  10. 4. Scientific Accomplishments � High Throughput Low Delay Network Sharing Bound Power Laws in Network Analysis 
 Fat tails, hard limits, thin layers Access Delay in IEEE 802.11 Stability of a P2P Protocol Target Tracking with In-Network Aggregation 
 Connection-Level Scheduling 
 E fg ect of Bu fg ers on Stability Security Cost of Network Security Reliability

  11. 4. Scientific Accomplishments � High Throughput Low Delay Network Sharing Bound Power Laws in Network Analysis 
 Fat tails, hard limits, thin layers Goal: Understand Capacity w. Network Coding Access Delay in IEEE 802.11 Stability of a P2P Protocol 
 Approach: Connection-Level Scheduling 
 o Unicast and multicast well-understood E fg ect of Bu fg ers on Stability o What about multiple unicast? Presentation by Barlas Oguz Security Cost of Network Security Reliability

  12. 4. Scientific Accomplishments � High Throughput Low Delay Network Sharing Bound Power Laws in Network Analysis 
 Fat tails, hard limits, thin layers Access Delay in IEEE 802.11 Goal: Estimation of Distribution with HT Stability of a P2P Protocol 
 Connection-Level Scheduling 
 E fg ect of Bu fg ers on Stability Approach: o Study distribution of flow sizes o Sampling central notes in large graphs 
 Security Cost of Network Security Reliability Presentation by Bruno Ribeiro

  13. 4. Scientific Accomplishments � High Throughput Low Delay Network Sharing Bound Power Laws in Network Analysis 
 Fat tails, hard limits, thin layers Access Delay in IEEE 802.11 Goal: Estimation with HT; Role of Layers Stability of a P2P Protocol 
 Connection-Level Scheduling 
 E fg ect of Bu fg ers on Stability Approach: o Goodness of fit with HT; o Hard limits; Security o Essentials of layered architecture. Cost of Network Security Reliability Presentation by John Doyle

  14. Goals: Access delay distribution; CW adaptation 4. Scientific Accomplishments � High Throughput/ Low Energy Approach: Low Delay o Distributed estimation of access delays o CW adaptation based on collision parameters, 
 Network Sharing Bound not recent collision history. Power Laws in Network Analysis 
 Presentation by Avideh Zakhor Fat tails, hard limits, thin layers Access Delay in IEEE 802.11 Stability of a P2P Protocol Target Tracking with In-Network Aggregation 
 Connection-Level Scheduling 
 E fg ect of Bu fg ers on Stability Security Cost of Network Security Reliability

  15. Goal: Stable distributed P2P protocol 4. Scientific Accomplishments � High Throughput/ Low Energy Approach: Low Delay o Avoid “rare chunk problem” by sampling. 
 Network Sharing Bound Presentation by Barlas Oguz Power Laws in Network Analysis 
 Fat tails, hard limits, thin layers Access Delay in IEEE 802.11 Stability of a P2P Protocol Target Tracking with In-Network Aggregation 
 Connection-Level Scheduling 
 E fg ect of Bu fg ers on Stability Security Cost of Network Security Reliability

  16. Goal: Reduce variance of target location estimate 4. Scientific Accomplishments � High Throughput/ Low Energy Approach: Low Delay o Aggregate information in network. 
 Network Sharing Bound Presentation by Wei Wei Power Laws in Network Analysis 
 Fat tails, hard limits, thin layers Access Delay in IEEE 802.11 Stability of a P2P Protocol Target Tracking with In-Network Aggregation 
 Connection-Level Scheduling 
 E fg ect of Bu fg ers on Stability Security Cost of Network Security Reliability

  17. 4. Scientific Accomplishments � Goal: Stability when connections come and go. High Throughput/ Low Energy Approach: Low Delay o MWM with suitable f(Q MAC ). 
 Network Sharing Bound Power Laws in Network Analysis 
 Presentation by Javad Ghaderi Fat tails, hard limits, thin layers Access Delay in IEEE 802.11 Stability of a P2P Protocol Target Tracking with In-Network Aggregation 
 Connection-Level Scheduling 
 E fg ect of Bu fg ers on Stability Security Cost of Network Security Reliability

  18. 4. Scientific Accomplishments � Goal: Understand stability of TCP with large bu fg ers. High Throughput/ Low Energy Low Delay Approach: o Understand conditions when TCP remains stable 
 Network Sharing Bound Power Laws in Network Analysis 
 Presentation by Somayeh Sojoudi Fat tails, hard limits, thin layers Access Delay in IEEE 802.11 Stability of a P2P Protocol Target Tracking with In-Network Aggregation 
 Connection-Level Scheduling 
 E fg ect of Bu fg ers on Stability Security Cost of Network Security Reliability

  19. 4. Scientific Accomplishments � High Throughput Low Delay Goal: Impact of connectivity value on network design. Network Sharing Bound Power Laws in Network Analysis 
 Fat tails, hard limits, thin layers Approach: o Define value of network and cost of connectivity loss. Access Delay and CW Adaptation in IEEE 802.11 o Study NE of strategic attack vs. proactive defense. 
 Stability of a P2P Protocol 
 Connection-Level Scheduling 
 E fg ect of Bu fg ers on Stability Presentation by Assane Gueye Security Cost of Network Security Reliability

  20. 5. Group Synergy � Collaboration UIUC - UCB Delays in Adaptive CSMA Protocol Collaboration UIUC - UMASS Ying – Srikant - Towlsey Collaboration UMASS - CALTECH File Partitioning Algorithms Collaboration UCB – CALTECH Libin Jiang (UCB PhD) joined Caltech (as postdoc) Collaboration across projects UMASS – Shroff: Mutlipath protocols and LRD Jiang – Modiano: HT and Q-CSMA Gueye – Baras: Security and Topologies Continued collaboration across projects Exploration of delay-based protocols Combination of LQF and Adaptive CSMA

  21. 6. Future Plans � 1) Analysis of delays and likelihood of delivery in wireless networks; 2) Design of network topologies for reliability and resiliency to attacks; 
 3) Improvements of WiFi protocols; 
 4) Study of LRD of functionals of Markov chains; 5) Development of secure wireless communication algorithms; 6) Design of resilient topologies

  22. 7. Original Goals � Improve Understanding of Multi-Scale Features of Networks Control protocols that capture the e fg ect of bu fg ers; Analysis of interaction of speed scaling and load balancing; Optimization of consensus algorithm; PCSDE models of heavy tails. Using that Understanding to Develop Improved Protocols Reduced-delay protocols for wireless networks; Stable distributed peer-to-peer algorithms; Stable multipath rate control algorithms; WiFi protocols with carrier sense and packet length adaptation; Geographical load balancing algorithms for renewables; Design of robust scheduling algorithms for heavy-tails; Algorithms for identifying important nodes in large networks; Algorithms for maintaining secure wireless communications.

  23. Today ʼ s Agenda � 9.00 - 9.40 Barlas Oguz (Berkeley) Network Sharing Bound and Stability of a P2P Protocol 9.40 - 10.20 Avideh Zakhor (Berkeley) Access Delay and CW Adaptation in IEEE 802.11 10.20 - 10.40 Break 10.40 - 11.20 Assane Gueye (NIST) Cost of Network Security 11.20 - 12.00 Javed Ghaderi (UIUC) Connection-Level Scheduling 12.00 - 1.00 Lunch Break 1.00 - 1.40 Wei Wei (UMASS) Target Tracking with In-network Aggregation 1.40 - 2.20 Bruno Ribeiro (UMASS) Power Laws in Network Analysis 2.20 - 3.00 Somayeh Sojoudi (Caltech) E fg ect of bu fg ers on stability of Internet congestion control 3.00 - 3.20 Break 3.20 - 4.00 John Doyle (Caltech) Fat tails, hard limits, thin layers 4.00 - 4.20 Feedback and Remarks from Dr. Chang

Recommend


More recommend