Ferry Replacement Protocols in Sparse MANET Message Ferrying Systems

1
Ferry Replacement Protocols in Sparse MANET
Message Ferrying Systems

• Jeonghwa Yang, Yang Chen,
• Mostafa Ammar, Chungki Lee
• Networking Telecommunications Group
• College of Computing
• Georgia Institute of Technology

IEEE Wireless Communications and Networking
Conference, 2005
2
Contents

• Background
• Message Ferrying System
• Problem Definition
• Ferry Replacement Protocols
• Concluding Remark

3
Background

• Ad hoc networks
• Connected networks
• Store-forward routing paradigm
• Most routing protocols
• Partitioned ad hoc network
• Wide deployment area, node mobility, limited
  radio range, physical obstacles
• Store-carry-forward routing paradigm

4
Message Ferrying System (cont.)

• Message ferrying system
• Developed by our group, Georgia Tech
• Designed as a routing strategy based on the
  store-carry-forward paradigm for sparsely
  distributed MANETs.
• A set of special nodes, called ferries, are
  responsible for delivering messages for all nodes
  in the networks.
• Exploits proactive mobility of ferries and nodes

W. Zhao, M. Ammar, and E. Zegura, A message
ferrying approach for data delay in sparse
mobile ad hoc networks, ACM Mobihoc, May 2004.
5
Message Ferrying System (cont.)
MF
6
Message Ferry Context

• A single ferry and a set of nodes are deployed.
• The ferry moves along the well-known ferry route
  at possibly variable speed.
• If nodes need to send or receive messages, they
  adapt their trajectory to meet the ferry.
• When the number of messages in a buffer reaches a
  certain threshold or when a node wants to poll
  the ferry
• Each node is associated with different amount of
  resources in terms of movement speed, buffer and
  battery, which is represented by real numbers
  called capability levels.

7
Ferry Replacement Protocols

• Intermittent but regular connectivity
• Howeversingle point of failure
• Graceful ferry stop vs. abrupt ferry failure

• Ferry replacement protocols in the presence of
  the abrupt ferry failure

8
Problem Definition

• Problem statement
• Design a protocol that makes one and only one
  ferry replacement substitute for the failed ferry
  after the abrupt ferry failure.
• Design considerations
• Recovery latency
• Protocol overhead on the ferry and nodes
• Success ratio of protocol

9
Ferry Replacement Protocols

• Successor designation protocol
• The responsibility to select a ferry replacement
  is on the current ferry.
• Distributed election protocol
• The responsibility to select a ferry replacement
  is distributed among nodes

10
Successor Designation Protocol

• Successor designation
• Current ferry
• succ_cl_threshold
• Redesignation in case of the successor
  failure
• Ferry failure detection
• Successor
• Ferry status check every chk_interval
• ferry_cycle_time
• Ferry replacement in case of the ferry
  failure

new ferry
CL gt succ_cl_threshold
successor
successor
successor
MF
MF
At time (t chk_interval)
At time (t)
11
Successor Designation Protocol (cont.)
Effect of the successors ferry status check
interval
Detection latency
Overhead
Ferry status check interval
Ferry status check interval
Movement overhead on successor and nodes
Failure detection latency
12
Distributed Election Protocol

• Ferry failure detection
• Potentially all nodes
• ferry_cycle_time
• Detector
• backoff_delay F(capability_level)
• Ferry election
• Candidate or elector
• capability_level
• highest_candidate (HC)
• WIN/LOSE message

9
backoff_delay
6
Elector1 HCc1(3)
3
backoff_delay
detector
candidate1
4
At time t1
13
Distributed Election Protocol

• Ferry failure detection
• Potentially all nodes
• ferry_cycle_time
• Detector
• backoff_delay F(capability_level)
• Ferry election
• Candidate or elector
• capability_level
• highest_candidate (HC)
• WIN/LOSE message

14
Distributed Election Protocol (cont.)

• Properties
• The election protocol ensures that exactly one
  node is elected as a ferry after no more than
  two rounds from the time the first node detects
  ferry failure under the condition that there is
  at least one elector.
• Proof
• Base step of candidates 1
• Induction of candidates k, k gt 1
• Election failure
• The election protocol fails when there is no
  elector forever

15
Distributed Election Protocol (cont.)
Effect of the backoff delay scale
Disrupted candidate ratio
Election latency
16
Concluding Remark

• Ferry replacement problem in the message ferrying
  system in sparse MANETs
• Two ferry replacement protocols
• Successor designation scheme
• Distributed election scheme
• Balance between performance and overhead by
  tuning protocol parameters
• Ferry replacement protocol in multiple ferry
  systems

17

• Thank You!!
• jeonghwa_at_cc.gatech.edu