Ad-hoc Storage Overlay System (ASOS): A Delay-Tolerant Approach in MANETs

Ad-hoc Storage Overlay System (ASOS): A Delay-Tolerant

Approach in MANETs

Guang Yang, Ling-Jyh Chen, Tony Sun, Biao Zhou and Mario Gerla

Computer Science Department, UCLA

Mobile Adhoc and Sensor Systems (MASS), 2006 IEEE International Conference on

Outline

Introduction Application Scenarios and Design

Principles The Ad-hoc Storage Overlay System

(ASOS) Architecture Probabilistic Data Replication Evaluation Conclusion

INTRODUCTION

MANETs set up rapidly without pre-deployed infras

tructure Military battlefields, disaster recovery

tasks must be fulfilled quickly Nodes may crash, lose power, be blocked,

or move out of the communication range of its neighbors

Motivation

End-to-end connectivity is highly susceptible to disruption Adjusting the motion of existing

nodes or deploying additional nodes can improve it

For scenarios where connectivity cannot be immediately improved

Proposed solution: Ad-hoc Storage Overlay System (ASOS)

Assumption A heterogenous MANET environment File transfer are considered as delay toler

ant A number of mobile nodes capture useful

data and send back to the control center Part of the data may not be delivered to t

he control center in time due to connectivity disruptions

Design Principles

Safe and robust storage Immediate availability Efficient storage and easy delivery Friendly interface

ASOS architecture

Initialization and Maintenance of ASOS Selecting ASOS peers Peer and file IDs Initialization Maintenance

ASOS Interface Data Management

ASOS architecture

source ASOS

disconnect

destination

Selecting ASOS peers

ASOS is a self-organized P2P overlay of existing nodes

It is desirable to designate a set of powerful nodes

Other regular nodes must understand the ASOS interface to access the storage utility

Peer and file IDs

Every node has a unique ID Files can be uniquely identified by

hashing from source/destination IDs and other information

Initialization

All ASOS peers form a multicast group Each ASOS peer multicasts periodic HELL

O messages to all ASOS peers Every peer can hear from all other reacha

ble peers and know which files are stored in ASOS and where they are

Format of a HELLO message

Maintenance

Each ASOS peer maintains a lookup table of reachable neighbors

Entries in the table contain similar fields as in the HELLO messages

Each entry is also associated with an expiration timer that is reset when the entry is refreshed

ASOS architecture

Initialization and Maintenance of ASOS ASOS Interface

Advertising of ASOS peers Disruption detection and data submission Data retrieval from ASOS

Data Management

Advertising of ASOS peers

Regular nodes could look at the HELLO messages to learn about the nearby ASOS peers and stored files Overhead

Introduce a new type of ADVERTISE messages

One node only forwards the first fresh ADVERTISE message it receives

Decouple the task of advertising

ASOS

A

ADVERTISE message

Entries

lookup table

A1

10unit(5,10)files

F

D

GC

E

B

a A 100k info

Disruption detection and data submission

ASOS

ASOS ASOS

A

Source

B

C

destination

disconnected

RERRRERR

Switching end-to-end flowto the ASOS mode

Disruption detection and data submission

ASOS

ASOS ASOS

A

Source

B

C

destination

disconnected

File Start Size

1 200 150

File Start Size

1 150 50

File Start Size

1 0 200

be divided into several chunks

Data retrieval from ASOS: pull model

ASOS

ASOS ASOS

A

Source

B

C

destinationADVERTISE

Data retrieval from ASOS: push model

ASOS

ASOS ASOS

A

Source

B

C

destination

hello

ASOS architecture Initialization and Maintenance of

ASOS ASOS Interface Data Management

Probabilistic selection of storage locations

Other data transfer between ASOS peers

Data deletion and replacement

Probabilistic selection of storage locations

To increase storage reliability, data is also replicated to other ASOS peers

Storage locations are selected by probabilistic replication: A peer closer to the destination node A peer further away from other ASOS

peers A peer less heavily loaded have a higher

probability to be selected

Other data transfer between ASOS peers

An ASOS peer may also dynamically transfer stored data to another peer running short of power or storage space

Data deletion and replacement

When the data is successfully delivered or has lost its usefulness

The original source or destination message its ASOS agent to delete data

The agent disseminates this message to all ASOS peers

ASOS Peer Deployment and Probabilistic Location selection under the VT Mobility Model

The Virtual Track (VT) mobility model targets the MANET scenario where mobility of the grouped nodes is constrained

Initially, each group contains a certain number of ASOS peers

Due to splitting, a group may temporarily have zero ASOS peers nodes turn to ASOS peers, or to static/individual

peers

The Virtual Track mobility model

Evaluation Application module: QualNet Simulation scenario: UCLA campus map 30 Nodes 1600 m×1600 m square area: 5 static,

5 individual (RWP) and 20 grouped (divided into 4 groups)

Each source generates a periodic constant-bit-rate (CBR) flow, at the rate of 80 Kbps, for 10 seconds every minute

Each simulation runs 20 minutes

Screen snapshot of the simulation topology in QualNet

Simulation parameters are summarized

Simulation parameters are summarized

Instantaneous throughput measured at the destination node

Cumulative amount of data delivered to the destination as time proceeds

Delivery ratios of data generated in each minute

static node individual node grouped node

Cumulative distribution function (CDF) of the size of HELLO messages

Delivery ratio vs. number of ASOS peers

Delivery ratio vs. number of replicated data copies

Conclusion

The Ad-hoc Storage Overlay System (ASOS) that extends end-to-end data transport in MANETs when connectivity is disrupted

Storing undeliverable data reliably in an overlay of storage-abundant

Implemented the ASOS in QualNet

Future work

ASOS peers can be dynamically elected Data encryption, user authentication and

intrusion detection are needed Soft disruptions

Thank you!