Application-aware in-network service deployment for collaborative adaptive sensing of the atmosphere (CASA)

(1)

Application-Aware In-Network Service Deployment for Collaborative

Adaptive Sensing of the Atmosphere (CASA)

)

Panho Lee, Tarun Banka, Sanghun Lim, Anura P. Jayasumana, V. Chandrasekar

{leepanho, banka, shlim, anura, chandra}@engr.colostate.edu

This work was supported primarily by the Engineering Research Centers program of the National Science Foundation under NSF award number 0313747. Any opinions, findings, conclusions, or recommendations expressed in this material are those of the authors and do not necessarily reflect those of the National Science Foundation.

Department of Electrical and Computer Engineering, Colorado State University, Fort Collins

CASA:

C

ollaborative

A

daptive

S

ensing of the

A

tmosphere

10,000 ft 10,000 ft tornado tornado wind wind earth_surf ace earth_surf ace snow snow 3 .0 5 k m 3 .0 5 k m 3.0 5 k m 3.0 5 k m 0 0 4040 8080 120120 160160 200200 240240 RANGE (km) RANGE (km)

Collaborating Collaborating radars:radars:

improved sensingimproved sensing

improved detection, prediction improved detection, prediction

Responsive to multiple Responsive to multiple endend--user needsuser needs 10,000 ft 10,000 ft tornado tornado wind wind earth_surf ace earth_surf ace snow snow 3 .0 5 k m 3 .0 5 k m 3.0 5 k m 3.0 5 k m 0 0 4040 8080 120120 160160 200200 240240 RANGE (km) RANGE (km) Horz. Scale: 1 Horz. Scale: 1””= 50 km= 50 km Vert. Scale: 1 Vert. Scale: 1””--==--2 km2 km 5.4 k m 5.4 k m 1 k m 1 k m 2 km 2 k m 4 k m 4 k m gap gap

Current State of the Art

Current State of the Art CASA VisionCASA Vision

Sparse, highSparse, high--power radarpower radar

Sensing gap: Earth curvature Sensing gap: Earth curvature effects prevent 72% of the

effects prevent 72% of the

troposphere below 1 km from

being observed

CASA Oklahoma Test

-

bed

System Environment

Initial 4Initial 4--nodes testnodes test--bed (Sensor network can be extended to tens of nodes)bed (Sensor network can be extended to tens of nodes)

Wired/wireless TCP/IP communicationWired/wireless TCP/IP communication

ResourceResource--rich sensing, processing and communicationrich sensing, processing and communication

End user requirements

Distinct bandwidth requirementsDistinct bandwidth requirements

Data quality requirementData quality requirement

Single/Single/Multi sensor algorithmsMulti sensor algorithms

Differing sensing requirementsDiffering sensing requirements

Radar 1 Radar 1 Radar 2 Radar 2 Radar 3 Radar 3 Radar 4 Radar 4

Multi

-

Sensor Data Fusion: Challenges

Peer

Peer--toto--Peer Processing:Peer Processing:

P2P for multiP2P for multi--sensor fusion sensor fusion

Better resource utilization by sharing Better resource utilization by sharing

Coordinated processingCoordinated processing

Better fault toleranceBetter fault tolerance

Coordinate peers to minimize response timeCoordinate peers to minimize response time Resource

Resource--intensive Applications:intensive Applications:

HighHigh--bandwidth data generation:bandwidth data generation:

several Mbps to tens of Mbps per nodeseveral Mbps to tens of Mbps per node

Vast amount of computation resourcesVast amount of computation resources

ClientClient--Server architecture may not be Server architecture may not be

appropriate:

Scalability IssueScalability Issue

SingleSingle--point of failurepoint of failure

Preprocessing Preprocessing Preprocessing Preprocessing Preprocessing Preprocessing Data Fusion Data Fusion Radar A Radar A Radar B Radar B Radar C Radar C Data Fusion Data Fusion Data Fusion Data Fusion

Fusion Node Architecture

Performance Evaluation

Abstract

An Engineering Research Center for Collaborative Adaptive Sensin

An Engineering Research Center for Collaborative Adaptive Sensing of the Atmosphere g of the Atmosphere

(CASA) funded by the National Science Foundation, seeks to revol

(CASA) funded by the National Science Foundation, seeks to revolutionize the way we detect, monitor utionize the way we detect, monitor

and predict atmospheric phenomena by creating a dense network of

and predict atmospheric phenomena by creating a dense network ofsmall, lowsmall, low--cost, lowcost, low--power radars power radars

that could collaboratively and adaptively sense the lower atmosp

that could collaboratively and adaptively sense the lower atmosphere. Such a network is expected to here. Such a network is expected to provide more timely and accurate forecasts for tornadoes, flash

provide more timely and accurate forecasts for tornadoes, flash floods, and other hazardous weathers. floods, and other hazardous weathers.

In addition, the networked radars can offer improved accuracies

In addition, the networked radars can offer improved accuracies and more specific inferences that and more specific inferences that

could not be achieved by the use of a single long

could not be achieved by the use of a single long--range radar. In CASA, multiple end users may be range radar. In CASA, multiple end users may be

present that have distinct sensing, communication and computatio

present that have distinct sensing, communication and computation requirements for their operations. n requirements for their operations.

In addition, the underlying network infrastructure may itself be

In addition, the underlying network infrastructure may itself besubjected to adverse conditions due to subjected to adverse conditions due to

severe weather and link degradation/outage along wired and wirel

severe weather and link degradation/outage along wired and wireless links. ess links.

We use overlay networking to provide acceptable quality of servi

We use overlay networking to provide acceptable quality of service (QoS) and robust data ce (QoS) and robust data

transport service for the CASA end

transport service for the CASA end--users. At CSU, we have developed an AWON (Applicationusers. At CSU, we have developed an AWON (Application--aWare aWare

Overlay Networks) architecture for deploying application

Overlay Networks) architecture for deploying application--aware services in an overlay network to best aware services in an overlay network to best

meet the end

meet the end--usersusers’’QoS requirements over the available networking infrastructure; QoS requirements over the available networking infrastructure; based on this, we based on this, we

have implemented an application

have implemented an application--aware multicast service for CASA. We also present a multiaware multicast service for CASA. We also present a multi--sensor sensor

fusion framework which can provide a mechanism for selecting a s

fusion framework which can provide a mechanism for selecting a set of data for data fusion considering et of data for data fusion considering

application

application--specific needs, and a distributed processing scheme to minimize specific needs, and a distributed processing scheme to minimize the execution time the execution time

required for processing data per integration algorithm.

P. Lee, T. Banka, A. P. Jayasumana, V. Chandrasekar, P. Lee, T. Banka, A. P. Jayasumana, V. Chandrasekar, ““ContentContent--based Packet Marking for Applicationbased Packet Marking for Application--aware Processing in Overlay Networks,aware Processing in Overlay Networks,””Proc. of IEEE Local Proc. of IEEE Local

Computer Networks LCN 2006, pp. 123

Computer Networks LCN 2006, pp. 123--131, Tampa, FL, Nov. 2006131, Tampa, FL, Nov. 2006

T. Banka, P. Lee, A. P. Jayasumana, and J.F. Kurose, T. Banka, P. Lee, A. P. Jayasumana, and J.F. Kurose, ““An Architecture and a Programming Interface for ApplicationAn Architecture and a Programming Interface for Application--Aware Data Dissemination Using Overlay Aware Data Dissemination Using Overlay

Networks,

Networks,””Proc. of IEEE/ACM 2nd Intl. Conf. on Communication System SoftwProc. of IEEE/ACM 2nd Intl. Conf. on Communication System Software and Middleware, COMSWARE, 2007, pp. 1are and Middleware, COMSWARE, 2007, pp. 1--11, Bangalore, India, Jan. 2007 11, Bangalore, India, Jan. 2007

P. Lee, A. P. Jayasumana,P. Lee, A. P. Jayasumana,S. Lim and V. Chandrasekar,S. Lim and V. Chandrasekar,"A Peer"A Peer--toto--Peer Collaboration Framework for Multisensor Data Fusion,"Peer Collaboration Framework for Multisensor Data Fusion,"Proc. International Joint Proc. International Joint

Conferences on Computer, Information, and Systems Sciences, and

Conferences on Computer, Information, and Systems Sciences, and Engineering (CISSE 2007),Engineering (CISSE 2007),Bridgeport, CT,Bridgeport, CT,Dec. 2007.Dec. 2007.

S. S. Lim, V. Chandrasekar, P. Lee, A. P. Jayasumana, "Reflectivity ReLim, V. Chandrasekar, P. Lee, A. P. Jayasumana, "Reflectivity Retrieval in a networked radar environment: Demonstration from thetrieval in a networked radar environment: Demonstration from theCASA IPCASA IP--1 radar network," 1 radar network,"

Proceedings of IGARSS07, Barcelona, Spain

Publications

Application

-

Aware Overlay Network

PACKET LOSS

Application

-

Aware Multicast Protocol

Data Selection Data Selection TRABOL Congestion Control

TRABOL Congestion Control ApplicationApplication--Aware MulticastAware Multicast

AWON (

A

pplication a

W

are

O

verlay

N

etwork) Node Architecture:

CASA Data Fusion

Network

Network--based Reflectivity Retrieval:based Reflectivity Retrieval:

Set of observations from multiple radars Set of observations from multiple radars

Improve sensing accuracy by combining data from multiple radarImprove sensing accuracy by combining data from multiple radarss Observation from Radar 1

Observation from Radar 1 Observation from Radar 2Observation from Radar 2

Observation from Radar 3

Observation from Radar 3 Observation from Radar 4Observation from Radar 4 Combined Multiple Radar Combined Multiple Radar Observation Observation 0.5 1.5 2.5 3.5 141142143144145146147148149150 Gate No. S ta n d a rd D e v ia ti o n o f R e fl e c ti v it y (d B z )

Random Drop, Random Forw arding Selec tiv e Drop, Random Forw arding Selec tiv e Drop, Forw arding w ith Packet Marking No loss

Application

Application--aware Multicastaware Multicast

Data Fusion Response time Data Fusion Response time

Application

-

Aware Multicast

Variable System Load Condition Variable System Load Condition

Develop networking protocols for emerging distributed collaborat

Develop networking protocols for emerging distributed collaborative adaptive sensing ive adaptive sensing

systems such as CASA:

Develop an architecture for applicationDevelop an architecture for application--aware data aware data

dissemination

Design and develop applicationDesign and develop application--aware protocolsaware protocols

Develop application programming interface(API) for Develop application programming interface(API) for

application

application--aware overlay deploymentaware overlay deployment

Develop a multiDevelop a multi--sensor data fusion framework for CASA sensor data fusion framework for CASA

radar data fusion algorithms

Goals

ApplicationApplication--aware processing paradigm using overlay networks for meeting QoSaware processing paradigm using overlay networks for meeting QoSrequirements of heterogeneous end users in CASA networksrequirements of heterogeneous end users in CASA networks

AWON architecture for deployment of applicationAWON architecture for deployment of application--aware functionality in overlay networksaware functionality in overlay networks

ApplicationApplication--aware congestion control and inaware congestion control and in--network processing enhance the QoS under dynamic network conditinetwork processing enhance the QoS under dynamic network conditionsons

MultiMulti--sensor data fusion framework for CASA and other emerging distribsensor data fusion framework for CASA and other emerging distributed collaborative adaptive sensing systemsuted collaborative adaptive sensing systems

Deployment of multiDeployment of multi--sensor data fusion service in CASA networksensor data fusion service in CASA network

Data fusion framework reduces the execution time required for dData fusion framework reduces the execution time required for data fusion processing ata fusion processing

IImplement a prototype of multimplement a prototype of multi--sensor data fusion framework and test it in a real testsensor data fusion framework and test it in a real test--bed environment. bed environment.

DDevelop an analytical model for data fusion response time consideevelop an analytical model for data fusion response time considering the impact of network dynamics such as network delay and loring the impact of network dynamics such as network delay and lossss