An Application-Aware Overlay Networks Architecture & API Tarun Banka, Panho Lee, H. M. N. Dilum Bandara, and Anura P. Jayasumana Department of Electrical and Computer Engineering, Colorado State University, Fort Collins, CO 80523. Dilum Bandara dilumb@engr.colostate.edu
Outline • CASA • Application aware networks • Overlay networks • AWON architecture & API • What’s ahead FRACTAL 2009 2
Collaborative Adaptive Sensing of the Atmosphere (CASA) • Concept 10,000 ft 3.05 km 4 km – A set of networked small radars 2 km 1 km instead of a large radar 5.4 km snow – Sense lower 3 km of atmosphere 3.05 km wind tornado gap e a – Collaborating radars: Horz. Scale: 1” = 50 km r t h s u Vert. Scale: 1” ‐=‐ 2 km r f a c e 0 40 80 120 160 200 240 • Improved sensing, detection, & RANGE (km) prediction 10,000 ft 3.05 km • Goal – Improve warning time & forecast snow 3.05 km accuracy for hazardous weather wind tornado 0 40 80 120 160 200 240 RANGE (km) FRACTAL 2009 3
CASA Oklahoma Test-bed DS-3 SONET WesternHeights 4 5 M Each Hub Has existing b - Fiber Based DS-3 to E t h OneNet e OneNethub-OKC r n DS-3 e Hub-OKC t Radar 1 OU Engineering Radar 4 8 0 2 . 1 1 B DS-3 Newcastle DS-3 Tower- Newcast DS-3 USAO Chickasha Tower Chickash Tower Lexington OneNet DS-3 DS-3 Hub- DS-3 Chick 3550 DS-3 Radar 3 DS-3 RushSprings Radar 2 Tower- RushSpri ngs 3 3 D S - - S D DS-3 Lawton Repeater D S - 3 Velma Radar 1 Radar 2 WOSC Altus Tower- Lawton Tower- Tower- Velma Snyder Altus 100Mb Tower- Cameron Snyder OneNet Radio tower OneNet Hub- 100 Mb Hub- Lawton OneNet Tower- Altus Hub- Ardmore Ardmore • Multiple high bandwidth streams • Real-time communication • Heterogeneous infrastructure & end users • Simultaneous observations by multiple radars • Multi-sensor data fusion Radar 3 Radar 4 • Hostile weather conditions FRACTAL 2009 4
Challenges • Distributed Collaborative Adaptive Sensing (DCAS) – Sensing, processing, & storage elements are interconnected via the Internet – High bandwidth data streams – Real-time communication – End users rely on different subsets of data • Adapting to varying network conditions • QoS perceived by end users depends on – Timeliness & usefulness of received data – Jitter, excessive delay, & random packet loss • End-host based correction/adaption is not effective FRACTAL 2009 5
Application Aware Networks (AANs) • Networks that implicitly or explicitly learn about the application characteristics & adapt their forwarding strategy to meet the application requirements in the best possible way • Possibilities – Application aware packet forwarding – In-network application aware processing & storage – Enhance the adaptability of applications – Enhanced QoS provided to end users • Current networks are not application aware – Lack of hardware & architectural support • Overlay networks are a viable solution FRACTAL 2009 6
Overlay Networks • A computer network built on top of another network – e.g., dial-up Internet, P2P • Application layer solution • Virtual links between overlay nodes • Overlay routing – Control path/link selection – QoS enhancement • Application-aware processing at intermediate nodes FRACTAL 2009 7
Application-Aware Overlay Networks (AWON) Architecture • An architectural framework to deploy AAN services – ASP - Application-specific Service Plug-in – ASPs inject application specific functionality into overlay nodes – API for easy deployment of ASPs • www.cnrl.colostate.edu/Projects/AWON/awon- api.0.1.tar.gz • ASPs regulate flow of data through overlay nodes – Taking application specific constrains into account – Extract, select, fuse, & repack data • Overlay routing protocol provide desired QoS support FRACTAL 2009 8
AWON (cont.) • Different AAN services • Application specific content delivery under varying network conditions • Better quality in content delivered to end users Measurements on PlanetLab FRACTAL 2009 9
What’s Ahead • Basic functions of overlay networks will eventually be migrated into network elements • Potential for in-network processing & storage – However, resources per application will still be limited • Renewed interest in AANs – Vendors are interested in architectural support • AWON architecture would be a suitable reference framework in realizing the AAN paradigm – Overlay networks, AANs P2P based in-network fusion • How to distribute ASPs? – Online installation, micro-programs in packets, etc. – Need programming & protocol description languages FRACTAL 2009 10
Recommend
More recommend
