Research Review Rudra Dutta, CSC, NCSU April, 2014 Networking Frontiers l Ubiquity – – – l Software – – – – l “Up the stack” phenomenon Down the stack, also – virtualization And sideways – policy, measurement, economics, analytics Democratization of networking functionality (chaotic?) Architecture – – l Mobility, Addressing, Binding Dependability, Confidentiality, Availability Cyber-physical computing, sensing/actuating All of the above Design tweaks won’t do it all Some work in above areas Copyright Rudra Dutta, NCSU, Spring 2014 Software Defined Networking SDN Agent Flow Table l SDN Agent Flow Table SDN Agent Flow Table SDN Agent Flow Table SDN Agent Flow Table Embodiment of policy box – – – Forwarding policy (routing), security policy, energy policy, economic policy, … Realized in hardware or software Or platform running software “policy programs” Copyright Rudra Dutta, NCSU, Spring 2014 Controller Federation SDN Agent Flow Table l SDN Agent Flow Table SDN Agent Flow Table SDN Agent Flow Table Different datapaths belong to different entities – l SDN Agent Flow Table Controlled by different controllers Flows route through multiple domains – – Controllers must coordinate among themselves Possibly exchange limited privileges on datapaths Copyright Rudra Dutta, NCSU, Spring 2014 Programming the Controller SDN Agent Flow Table l l l SDN Agent Flow Table SDN Agent Flow Table SDN Agent Flow Table SDN Agent Flow Table Controller application must be written in some language – northbound API of controller Not well standardized in OpenFlow OpenDaylight – a proposal by an industry consortium – – Multiple southbound API support Standardized single northbound API Copyright Rudra Dutta, NCSU, Spring 2014 Network Virtualization (Many Faces of) l No “added functionality”, but “moving around” – – l An attempt to define complex network functions as abstractions: NF à VNF – – – l Makes functionality available where it was not An architectural distinction That can then be implemented to produce similar behavior on diverse platforms Orchestrated without knowing underlying platform Possibly as software, on VMs à NFaaS “Software Networking” ? Copyright Rudra Dutta, NCSU, Spring 2014 SDN Lab Copyright Rudra Dutta, NCSU, Spring 2014 Experiment – Traffic Flow: Best-Effort Model Control Plane Controller RTSP traffic faces delay as it goes through Traffic Controller 1 Traffic Controller 2 does NOT introduce any delay OpenFlow Datapath Traffic Controller 1 PC 1 EPB Checks EPB intercepts whetherClient this packet ports to allow access belongs to anfor expedited future traffic patterns ifand stream, so tells savesthe theDatapath URL associated to expediteto otherwise this flow.tells it use best-effort Copyright Rudra Dutta, NCSU, Spring 2014 Traffic Controller 2 Best-effort Route Expedited Route Data Plane Sending PC 1 video after replying to this message Legend Multimedia Server Network Management Evolution l Networking management processes need study – Especially in light of the coming SDN revolutio Processes are human processes - not crisp and clear l Data-driven = human-centered, here l Goal: use analytics to spot vulnerable process points, opportunities l – – Automation of SDN management Vulnerability elimination Copyright Rudra Dutta, NCSU, Spring 2014 What is CentMesh ? CentMesh : a project to build a versatile, open source shared resource on campus l Provide an active testbed to support research and instruction on wireless networks l Part of NCSU’s “Living Laboratory” vision l Copyright Rudra Dutta, NCSU, Spring 2014 Who Built CentMesh ? l Homemade ! – – – – Built by NCSU student researchers Guided by NCSU faculty Funded by Army Research Office and NCSU Operationalized by Institute of Technology of Next Generation (ITng), NCSU Copyright Rudra Dutta, NCSU, Spring 2014 Impact of Power Control in Mesh Copyright Rudra Dutta, NCSU, Spring 2014 Geodiffuse Routing A form of limited flooding l Nodes over an area form a diffuse path l Ready to forward transmission if most direct path seems to have difficulties l Can coordinate with heuristic mechanisms, and fine tune l Copyright Rudra Dutta, NCSU, Spring 2014 CentMesh R Kasyap Marmavula, Vaidy Ananthanarayan, Dr. Mihail Sichitiu, Dr. Rudra Dutta The NCSU Centennial Wireless Mesh Project Drones Challenge 2014 Hmm..what can you do? Can you traverse a path? Certainly – I can even avoid obstacles ! Take sensor readings or video, stream it back ? Yes – I tie into the NCSU CentMesh sesnornet ! Where did they get you from ?! NCSU faculty and students fabricated me – and put in a computer to make me autonomous ! (Thanks) WiFI BBBK CentMesh Mobile Node MAVLink over USB Abort Signal Autopilot So can I use you for teaching course XYZ or research ? I don’t know what that is but I can if you program me to do it ! CentMesh Drone PWM Commands Airframe CentMesh Mobile Node MAVLink over USB MCP over UDP/IP Autopilot CentMesh Drone Come fly with me ! Join the programming challenge go.ncsu.edu/drones_challenge_2014 Qualifying challenges: March 31st, online Final challenge: April 12th, Centennial Oval Drones Challenge 2014 Grand Sponsor: Aruba Networks CentMesh funded by: Army Research Office Copyright Rudra Dutta, NCSU, Spring 2014 Network Innovation through Choice Stupid computer. network. l l Informed exercise of choice (backed by money) can reward providers with good performance Select for helpful providers, beneficial ecosystem Copyright Rudra Dutta, NCSU, Spring 2014 Architectural Need Encourage Alternatives Vote with Your Wallet l l Know What Happened Informed exercise of choice (backed by money) can reward providers with good performance Select for helpful providers, beneficial ecosystem Copyright Rudra Dutta, NCSU, Spring 2014 ChoiceNet Entities / Interactions l Consumer – willing to exchange consideration for services deemed of value 😊 Provider – provides services in exchange of – l User who exercises choice à “chuser” consideration – – – – 😃 HW/SW infrastructure provider (path service) Marketplace provider Composition provider Verification provider Copyright Rudra Dutta, NCSU, Spring 2014 Entities and Interactions -------------Marketplace Search 😊 Customer Planner Help Plan ------- -- ------- ------- -- ------- ------- -- ------- ------- Advertise Measurements Verify Use l ------- Purchase Analyzer l Search -- Provision Monitor 👀 😃 Provider Service Informed exercise of choice (backed by money) can reward providers with good performance Encourage Alternatives Select for helpful providers, beneficial ecosystem Vote with Your Wallet Copyright Rudra Dutta, NCSU, Spring 2014 Know What Happened Jitter Apportionment for Video UX Raw MD Figures Time (s) 30 Time (s) 2.4 25 2.2 20 SP 2.0 MP1 MP2 1.8 Chooser 15 10 1.6 5 1.4 0 0 20 40 60 Marker Sequence Number Copyright Rudra Dutta, NCSU, Spring 2014 80 100 0 120 Close-up 1 2 Marker Sequence Number 3 Jitter Apportionment for Video UX Delay (ms) 700 600 mp1 owd 500 mp2 owd 400 Chooser owd Ji?er at Chooser 300 200 100 0 0 2 4 Copyright Rudra Dutta, NCSU, Spring 2014 6 8 10 12 14 16 18 20 22 24 26 28 30 Time (s) Jitter Apportionment for Video UX A freeze A glitch Delay (ms) 700 600 mp1 owd 500 mp2 owd Chooser owd 400 Glitch Freeze 300 Ji?er at Chooser 200 100 0 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 Time (s) Glitches denote the losses and Freezes denote video playback freezes Copyright Rudra Dutta, NCSU, Spring 2014 Jitter Apportionment for Video UX Delay (ms) Contributing One-Way-Delay per Provider 600 MP1 OWD No contributing delay at NSP2 500 400 MP2 OWD -‐ MP1 OWD 300 200 Chooser OWD -‐ MP2 OWD 100 0 4 8 10 12 15 17 20 22 25 28 Basic Analysis Results Provider NSP1 NSP2 NSP3 Copyright Rudra Dutta, NCSU, Spring 2014 Mean JiAer % 44.6 % 0.2 % 55.2 % Std. Dev. JiAer % 32.8 % 0.3 % 66.9 % Max JiAer % 25.3 % 0.2 % 74.5 % Time (s) Jitter Apportionment for Video UX Analysis for freeze 1 [t=6 and t=10] Mean % Std dev % NSP1 96.7 % Max J % 93.8 % 94.7 % NSP2 0.2 % 0.6 % 0.2 % NSP3 3.1 % 5.6 % 5.1 % Freeze 1 is at t=8s Delay (ms) 250 Freeze 2 is at t=10s 200 mp1 owd 150 Analysis for freeze 2 [t=8 and t=12] Mean % Std dev % NSP1 98.5 % mp2 owd Chooser owd 100 Glitch Max J % 97.2 % 97.2 % NSP2 0.0 % 0.0 % 0.0 % NSP3 1.5 % 2.8 % 2.8 % Freeze 50 Ji?er at Chooser 0 5 Copyright Rudra Dutta, NCSU, Spring 2014 7 9 11 Time (s) Jitter Apportionment for Video UX Freeze 3 is at t=19s Delay (ms) Analysis for freeze 3 [t=17 and t=21] Mean % Std dev % Max J % NSP1 27.7 % NSP2 0.2 % 42.2 % 50.2 % 0.4 % 0.3 % NSP3 72.1 % 57.4 % 49.5 % 700 600 Freeze 4 is at t=26s 500 mp1 owd mp2 owd 400 Chooser owd Analysis for freeze 4 [t=24 and t=28] Mean % Std dev % Max J % NSP1 NSP2 14.0 % 13.8 % 13.6 % 0.2 % 0.2 % 0.2 % 86.0 % 86.2 % NSP3 85.8 % Copyright Rudra Dutta, NCSU, Spring 2014 300 Glitch 200 Freeze 100 Ji?er at Chooser 0 16 18 20 22 24 26 28 Time (s) Security in the Cloud – Cost Points Customer Premise Customer Property Copyright Rudra Dutta, NCSU, Spring 2014 Industry Research Relationships l Challenges – Industry needs applicable research partnerships l Good deal of specificity l Time and effort to acclimatize – Faculty and students need to do research l Discovery of new knowledge, all that l Publish papers, pass PhD milestones – three years or so l Opportunities – – – – Computing is roaming free in the world Real-world problems, like it or not New problems, like it or not Synergy Copyright Rudra Dutta, NCSU, Spring 2014
© Copyright 2024 ExpyDoc
