(MCONF) for Tactical MANETs

Design and Evaluation of a Mass Configuration Protocol (MCONF) for Tactical MANETs Peter Katlic

Koushik Kar

James Nguyen

James Morris

Robert Cole

October 28, 2015

Outline

1. Introduction 2. Background and Related Work 3. Methodology 4. Findings 5. Discussion and Conclusion

2/35

Overview

Idea Tactical mobile ad-hoc networks (MANETs) can benefit from the ability to reliably execute a critical configuration change for all nodes. Must be able to: I

propagate in controlled manner

I

avoid node isolation

I

maintain network performance

3/35

Tactical Network

4/35

Challenges Lack of infrastructure I

nodes act as routers

I

multi-hop communications

I

control traffic overhead

I

partitions

Wireless medium I

loss

I

hidden/exposed terminal

I

overlap

I

contention

I

collision

5/35

Goal

Reliable high execution rate

Fast low duration

Efficient low number of sent messages

6/35

Outline

1. Introduction 2. Background and Related Work 3. Methodology 4. Findings 5. Discussion and Conclusion

7/35

Reliable Broadcast

I

Commonly used in ad-hoc wireless networks

I

Broadcast storm problem Approaches

I

I I I I I I I

reverse path forwarding (RPF) hot potato forwarding spanning tree dynamic probabilistic broadcasting double-covered broadcast counter/location/cluster-based simultaneous tx using frequency division

8/35

Configuration Protocols

I

Clock synchronization

I

Addressing (e.g. MANETconf)

I

Full configuration

I

Designed for wired networks (e.g. Network Configuration Protocol (NETCONF)) Approaches

I

I I I I

command-oriented variable-oriented object-oriented document-oriented

9/35

Consensus Protocols

I I

Focus of MANET research Requirements I I I

I I

agreement termination validity

Account for various network states Approaches I I I I

Hierarchical clusterer (HC) Proactive dissemination protocol (PDP) Reactive dissemination protocol (RDP) Proactive knowledge and reactive message (PKRM)

10/35

Simulation and Emulation

Advantages

Disadvantages

I

No physical devices

I

Time delays

I

Unmodified code

I

I

Scalable

Inaccuracy with high CPU load

I

Controlled

I

I

Repeatable

High performance computing for large networks

11/35

Outline

1. Introduction 2. Background and Related Work 3. Methodology 4. Findings 5. Discussion and Conclusion

12/35

MCONF Requirements

I

Propagate configuration change through entire MANET

I

Initiated by root node

I

Various modes of operation

I

Configurable

I

Decision process

I

Support for multiple routing protocols

13/35

Message Types

Type CONFIRM ACK COMMIT DONE LIST

Purpose Gather feedback or warn of impending configuration Respond with feedback and acknowledge impending configuration Configuration command Acknowledge configuration command execution Distribute known node addresses

14/35

Modes of Operation

Mode Baseline u-MCONF c-MCONF I

Early execution

I

Lazy execution

Description Single broadcast Unconditional with confirmation Conditional with confirmation

15/35

Baseline

Single broadcast I

controlled broadcast flood → baseline

I

safe and small changes

I

stable and connected network

16/35

u-MCONF

Unconditional with confirmation I

execution feedback

I

critical change without node input

I

tolerant of network disruption

17/35

u-MCONF Flowchart Start

Wait for Message

no

yes Root Node?

Broadcast COMMIT

yes

Wait for Message no

COMMIT? timeout yes Broadcast COMMIT to 1-hop Neighbors

no

DONE?

yes

Retry? yes

no

1-hop neighbors? no

DONE from all 1-hop neighbors?

yes yes Wait for Message

Execute Action

timeout no

DONE? Retry? yes

no

DONE from all 1-hop neighbors?

yes

Broadcast DONE

no

Execute Action

Stop

18/35

no

c-MCONF

Conditional with confirmation I

input gathering with execution feedback

I

critical change requiring node input

I

tolerant of network disruption

19/35

c-MCONF Flowchart (1/2) Start

Wait for Message

no

no

yes Root Node?

Broadcast CONFIRM

yes

Wait for Message

CONFIRM?

timeout

yes Broadcast CONFIRM to 1-hop Neighbors

yes

no

ACK? Retry? yes

no

1-hop neighbors? no

ACK from all 1-hop neighbors?

yes yes

Wait for Message

Make Decision

timeout

no

ACK? Retry? yes

no

ACK from all 1-hop neighbors?

yes

Broadcast ACK

no

20/35

no

c-MCONF Flowchart (2/2) no

Wait for Message

Execute?

timeout

yes no

COMMIT? Broadcast COMMIT

yes

yes Broadcast COMMIT to 1-hop Neighbors

Wait for Message

yes

timeout

no

1-hop neighbors?

no

DONE? Retry? yes

yes Wait for Message no

DONE from all 1-hop neighbors?

timeout

no

yes

DONE? Retry?

Execute Action

yes

no

DONE from all 1-hop neighbors?

yes

Broadcast DONE

no

Execute Action

Stop

21/35

no

Strategy Realistic Topologies I

Fixed

I

Mobile

MANET Emulation I

Multiple nodes with single machine

I

Repeatable and controlled conditions

I

Mobility Scenarios

Experimentation I

Modes

I

Message Sizes

I

Loss Rates

22/35

Assumptions

I

Addressing previously configured

I

Routing protocol in place

I

Root has one-hop neighbors

I

Mostly connected topology

I

Security/encryption ignored

23/35

Development and Testing Environment

I

Linux

I

LXC Linux containers

I

Extendable Mobile Ad-hoc Network Emulator (EMANE)

I

Common Open Research Emulator (CORE)

I

BonnMotion

I

Shell script

I

R

24/35

Outline

1. Introduction 2. Background and Related Work 3. Methodology 4. Findings 5. Discussion and Conclusion

25/35

Fixed Topology Details

I

10 nodes

I

node 10 as root

I

IEEE 802.11b/g 54mbps

I

100 replications

I

All modes of operation

I

Early and lazy execution

I

3 messages sizes (1KB, 5KB, 10KB)

I

5 loss rates (0-20%)

26/35

Average Early Execution Results - Small Messages

27/35

Average Early Execution Results - Large Messages

28/35

Average Early vs. Lazy Execution Results

29/35

Random Waypoint Topology Details

I

10 nodes move toward random points

I

stationary node 10 as root

I

RF ≈10mi range 54mbps

I

10 replications

I

All modes of operation

I

Early and lazy execution

I

3 messages sizes

I

range-based loss

I

mobility script

30/35

Random Waypoint Topology Results

Baseline u-MCONF (early exec) u-MCONF (lazy exec) c-MCONF (early exec) c-MCONF (lazy exec)

Message Size S M L S M L S M L S M L S M L

Execution Rate (%) 100.0 100.0 91.0 100.0 100.0 99.5 100.0 100.0 95.0 100.0 99.5 96.0 100.0 99.0 95.0

Confirmation Rate (%) 0.0 0.0 0.0 99.0 100.0 96.5 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 98.9

31/35

Duration (ms) 1198 1263 1238 1216 1627 2227 5285 5637 6945 2511 3869 4544 6719 7338 9984

Sent Messages 10.0 10.0 9.1 40.6 39.7 40.6 34.3 34.5 32.5 59.8 64.7 58.9 55.4 52.5 57.4

Outline

1. Introduction 2. Background and Related Work 3. Methodology 4. Findings 5. Discussion and Conclusion

32/35

Comparison to Existing Protocols

I

Broadcast flood I

I

I

I

similar ideas

Configuration I I

I

more robust

Reliable Broadcast

designed for MANETs confirmation for critical changes

Consensus I I

single deciding node useful approaches

33/35

Concluding Remarks

I

Successful execution of mass configuration change I I I

Confirmation Retransmission Execution tracking

I

Balance between configuration/consensus

I

Tunable

I

Routing protocol agnostic

I

Emulation shows promise for MANET protocol research

34/35

Thank You

Questions?

35/35