Experiences with Ad Hoc and P2P Networking Dr. Milena Radenkovic PhD, Computer Science, Nottingham, UK Dipl Ing., MSc, Electrical and Electronic Engineering,

Experiences with Ad Hoc and P2P Networking

Dr. Milena RadenkovicPhD, Computer Science, Nottingham, UK

Dipl Ing., MSc, Electrical and Electronic Engineering, Nis, SerbiaBA in Computer Science, American University in Bulgaria

University of Nottingham

Research Overview

Research interests centre on self-organised network architectures that support interactive multiuser applications in unstable and heterogeneous environments

Particular concern is with the design and deployment of novel mobile ad-hoc, delay tolerant network and P2P architectures for data store and query and routing protocols

Multiple applications in location based pervasive gaming, wearable medical and veterinary applications and mass environmental monitoring:

A Novel Routing Protocol for Large Scale Disconnected (PI, EPSRC),

Developing Advanced Collaborative Environments for Life Science Community (PI, EPSRC)

Participate (CI, EPSRC), myGrid (CI, EPSRC) IPERG, (WPLead, EU), MIAS (WPLead, EPSRC)


Supporting Collaborative Applications(Games) (past & current projects + PhD)

Collaborative Virtual Environments (CVEs), Inhabited TV, Mixed Reality Environments, (e.g. games and TV shows):

Merging physical and virtual worlds in mobile wireless systems

Extreme end system and network heterogeneity: convergence of media

Large number of simultaneous users and simultaneous live A/V streams but also heterogeneous sensor data

PhD concerned with the challenges for scaleable network multimedia architectures


Remote sensing with e-Science

Focus on science in the field: Scientists remote from the grid Access to grid facilities when

connections are limited or occasional

Environmental e-Science In Antarctic

Ice thickness, light intensity, water saltiness, GPS, weather info

and in the city CO and pedestrian movement

Remote Medical Monitoring Supplement medical

information respiratory, heart, blood,

temperature with physical information

Movement Location Posture

Davis Station


Distributed Partial Mixing (DPM) Approach (International Patent (WO 02/17579 A1) 2002, ACM Multimedia 2002, IEEE

Telsiks 2001, MIT Press PRESENCE, 2004)

Support for: the most natural and spontaneous verbal

communication packet loss management and delays maximal flexibility for audio tailoring at the end users heterogeneous networks and machines efficient distribution of audio/video streams in the

network congestion control: adaptability, inter-protocol fairness

LAN 2

i

j

Streams:a, b, c,d + e,f + g + h

LAN 1

a

b

e

d

c PMA

PMD

PMB

Streams:a, b, c,d + e

Streams:a, b, c, d, e


PMC

Dial-up Users

f

g

h

Streams:f + g + h

PME

LAN 2

i

j


LAN 1

a

b

e

d

c PMA

PMD

PMD

PMB

PMB



Streams:a, b, c, d, eStreams:a, b, c, d, e



PMC

Dial-up Users

f

g

h

Streams:f + g + hStreams:f + g + h

PME

Network driven

Application driven End system

resource driven


C6

C7

C3

C2C1New group

Congested link

C8

C4 C5

New group

DPM

DPM

Clients

Routers

Network Links

Self-organised, ad-hoc, fully distributed DPM

(ACM VRST 2002, MIT Press PRESENCE 2004)

Unmanaged deployment domain: frequent and severe fluctuations in link quality

DPM placed in end systems On-demand DPM initiation Distributed collaborative

controlling process: nodes continuously evaluate contribution of each node

Adaptive topology: congruent with the underlying physical network


PARTICIPATE (CI)

Continues to explore convergence in pervasive, online and broadcast media to create new kinds of mass-participatory events in which the public contributes to and accesses contextual content.

Aims to support participants’ understanding of individual impact on the environment AND to support change.

Develop an architecture that integrates diverse media, devices and networks

My Focus: How to support mass-scale sensor data capture and query

in challenged environments Investigate novel overlays that can help build applications

on the top of mobile, unstable and unreliable environments How do we support the querying of partially defined data

(i.e. range search queries)?



ScienceScope sensor and data logger, with screen shots from the

GeoMobSens application (collecting environmental audio and carbon monoxide levels).


Collected sound levels visualised on Google Earth


Carbon monoxide levels from the collected data


Mass-scale sensor data capture, route and query in challenged environments (ASWN ‘06, WONS ‘07, ICN ’08,

ICN ’09, WINET 09)

Novel locality preserving DHT overlays optimised for mobility and disconnections

New policies for forwarding and replication that utilise heterogeneity

Opportunistic utilization of locally available resources Novel DHT substrates (Layer 3 routing) optimised

for mobility and disconnections Utilises self organised caching and mobility to improve

dissemination of data (e.g. compared to other substrates) Novel mass Scale DTN Routing

Novel hybrid MANET and DTN routing protocols with DHT naming scheme

Use Bluetooth logs to indentify contact opportunities and mobility patterns in order to discover groups and profiles of users to design optimal forwarding and routing policies


A Possible Deployment Scenario

DHT overlay acrosspublicly owned infrastructure

DTN

MANET

Game Logic

WiFi

P2P Node

PSN

Mobile Phone

ContextPhone

Game Logic


7


“Time zones” of Friend, Acquaintance and Stranger Contact Clusters (digital footprint 1)

Node with the majority of “Friend” contacts considered more reliable and good to be a “data mule” or gateway


“Time zones” of Friend, Acquaintance and Stranger Contact Clusters (digital footprint 2)

Node with the majority of “Stranger” contacts suitable “ for data dissemination to a wide community


Cost Effective Mass Games in Unstable Environments

(WONS 2007)

Hybrid Pastry Exploits locality and high

knowledge about neighbours

3 types of nodes to account for heterogeneity Regular Clients (RCs) Super Clients (SCs) Application Servers (ASs)

Hybrid Pastry outperforms Pastry in terms of SR for average and low densities. Average delays are not increased.

DHT Overlay

MANET 2

MANET 1

Internet

Super Client 1

Regular Client

Regular Client

Super Client 2

Application Server

Super Client 3

Forwarding

Replication

Store(9bd546)

Application Server

Super Client

Regular Client

Query Forwarding

Comparing

1ac678

Get(9b3453)

951d57

9bc670

9b3450

9b3453

1ac678

Get(9bd546)

951d57

9bc670

9bd7219bd546

bd513a

Data Retrieval


DHT substrate to address disconnections(ASWN 2006, IEEE Multimedia 2008)

Mobile Ad-hocNetwork

Peer to Peer Internet Overlay

Sensor Node

Farm Building

Gateway Node


Farm Building

Pasture

Remote User

More Powerful Node Storing

Historical Data


Peer to Peer Internet Overlay

Sensor Node

Farm Building

Gateway Node


Farm Building

Pasture

Remote User

More Powerful Node Storing

Historical Data

State Contains data related to

animal’s health and location Is identified by animal’s id and

timestamp Every node stores its own state

and up to k states from other nodes

Results Success ratio twice as high

as EKTA’s in the case of increased mobility and disconnections while not increasing network overhead


Delay Tolerant Mass Scale Cattle Monitoring

(ICNS 2006, ICN 2007, ICN 2008, WiNET 2009)

Identify realistic requirements Strong demands for detecting

oestrus, lameness, animal diseases

Trails with bluetooth GPS, mobile phones, video footage

Enable continuous and delay tolerant multi-dimensional parameter monitoring

Utilises existing infrastructure but also works in fully ad-hoc mode

Retention of data, in-situ and remote queries and notifications

Increases fairness and decreases energy utilization

User

Pasture

Animal

AnimalMounted

Device (Collar)

Bu

ilt-i

nA

cce

lero

me

terA

nim

al A

rea

N

etw

ork

Multi Hop Wireless

Communication

Mu

lti H

op

W

ire

less

C

om

mu

nic

atio

n

0UserUser

Farm

Not

ifica

tions

an

d Q

uery

ing

FarmServer

BusinessCounterparts

Selective Data Sharing

UserUserNotifications

and Querying

Wired Network or GSM

Ga

tew

ay

Inte

rne

t

Ga

tew

ay

Ga

tew

ay

UserUserUser

Inte

rna

l S

tora

ge

a

nd

P

roce

ssin

g

Leg Mounted Pedometer

Pasture 2

Wired NetworkConnection

Pasture 3

GSM Connection

Pasture 1

Wired NetworkConnection

DSLConnection

Message Board

Terminal Server

Farm 1Farm 2

Internet

GSMConnection

Internet User

Mobile User

Gateway

GPRS modem

Pasture 4

MobileUser

MobileUser


Energy Efficient Route Discovery(ICN 2007, ICN 2008, WINET 2009)

Low data traffic but unstable path Save energy on path discovery (i.e. having more stable paths)

Minimizing and balancing energy utilisation Adaptive control of transmitter power Passive Clustering with Delayed intelligence (flood control) Selecting routes with longest lifetime

Least number of hops, Least number of deteriorating links, Minimal total power

Delaying non urgent data exchanges waiting for acquisition of a valid route to a target node from overheard or forwarded packets

Using movement patterns less mobile => more stable paths => more forward faster animals => better message carriers

Significant energy saving compared to ESDSR, more fair energy consumption and shorter delays


myGrid(CI), (ICWI 2003, CCGRID 2003, GADA 2005, SAG 2004, AHM 2005)

Open Source Upper Middleware for Bioinformatics

Developing high level services for data intensive integration rather than computationally intensive problems Workflows, Provenance Semantic-based resource discovery and

matching. My focus on the deployment perspective

Personalised semantic driven notification service Semantic driven web portal and gateway Integration with WSRF P2P Access Grid


Data at the centre

Provenance record of workflow run that produced this data

Provenance of the data

Workflows that could use this data

People who have registered an interest in this data

Ontologies describing data

Services that can use or produce this data

Notes

Data

Literature relevant

Literature relevant

Related Data


myGrid Services

#

Communication fabric

Text Extraction ServiceAMBIT

Workflow enactment engine

Distributed Query Processor

Provenance mgt

Personalisation

Event Notification

Gateway

Service and WorkflowDiscovery myGrid

Information Repository

Ontology Mgt

Metadata Mgt

Work benchTaverna

workflow environment Talisman application

Bio Services

Soaplab

Portal

Bio Services


myGrid deployment perspective

Web browser HTTP

XML-Protocol Gateway interface

Web Portal

Client libraries

XML-Protocol, CORBA, …

Client libraries

Wireless proxy

Client libraries

Handheld wireless device

Custom protocol(s)

PSE/IDE

Client libraries

Custom clients: desktop,

immersive

Client libraries

e-Science gateway

User proxy

Client libraries

XML-Protocol User proxy interface

Notification Service

Workflow Enactment

Service

Ontology Service

MyGrid Information Repository

service

Bioinformatics Services


Future research

Infrastructure and protocols that support true convergence of pervasive, online and broadcast media on the global scale

Embrace “radically heterogeneous networking” Two end-systems should be able to communicate even if they have nothing in common: protocols, address realms, semantics (Jon Crowcroft)

P2P, DTNs, PSNs but learn from Grids and WSs

Proposals in preparation: Trust, Reputation and support for security in disconnected worlds

Self organized security in DTNs in collaboration with Watson IBM Aim to extend DTN security bundle with novel flexible and fluid trust

building, negotiation and propagation mechanisms based on behavioral modeling, anomalous behavior across disconnections and non consensus asynchronous partial trust claiming and resolving

Congestion Control and QoS in disconnected worlds Self organized congestion control in DTNs in collaboration with Watson IBM Source driven and receiver driven feedback and rate adaptation

Application domains Medical, Military, Environmental, Social


Observed Movement Patterns


Realistic Parameters

Communication parameters Infrequent in-situ queries (field trials) Delivering data to sinks every 2-4 hours (field

trials) Low amount of data (literature)

Movement patterns Typical speed 0.6 m/s (literature) Rarely faster than 0.8 m/s (field trials) Stocking density on pastures 2-7 animals/ha

(literature) Drinking, eating and milking patterns (field trials) Movement patterns difficult to predict (field trials)


Simulation Results

Beacons, sinks. selecting routes with min.transmitter power requirements considerably decrease energy usage of mobile nodes during data transfers compared to ESDSR

At the cost of the increasing the delays of delivering data to sinks and energy usage of sinks


Simulation Results

Energy consumption of the new approach up to 40% lower than in the old approach

New approach offers more fair energy consumption and shorter delays


This time its personal

my services my favourite services my opinion of those services my workflows my data my notes my queries my logs of what I did The events I care about


Notification & Personalisation

Has PDB changed since I last ran this?

Has the record I derived my record from changed?

Has the workflow I adapted my workflow from changed?

Did the provenance record change?

Has a service I am using right now gone? Has an equivalent one sprung up?

Event notification service.

Dynamic creation of personal data sets in mIR

Personal views over repositories.

Personalisation of workflows.

Personal notification Annotation of datasets and

workflows. Personalised service

registries – what I think the service does, which services can GSK employees use


Simulation Results (Hybrid Pastry)


Simulation Results (DHT substrate)

Passive caching Always increases success

ratio (as expected) Decreases network overhead

in sparse topologies (better results than expected)

Increases scalability Proactive caching

Appropriate in case of very high mobility

Anticipated long periods of limited or no query traffic.

Does not increase success ratio as much as expected

Literature informed simulation parameters (no. of nodes, velocities)

Experiences with Ad Hoc and P2P Networking Dr. Milena Radenkovic PhD, Computer Science, Nottingham, UK Dipl Ing., MSc, Electrical and Electronic Engineering,

Documents

computer science

network heterogeneity

tolerant network

escience focus

life science community

novel routing protocol

veterinary applications

epsrcparticipate ci