course project book (mobile computing and wireless networks) cs 395t – spring 2001 course...
TRANSCRIPT
Course Project Book(Mobile Computing and Wireless Networks)
CS 395T – Spring 2001
Course Instructor:Dr. Yongguang Zhang ([email protected])
Course URL: http://www.cs.utexas.edu/users/ygz/395T
May 16, 2001
Project List
1. Mobile IP service for ACES and TAY2. Location Service for Mobile Device3. Linux Kernel Support for Ad-hoc Routing4. Linux Kernel Module for Ad-hoc Routing5. Multi-Mode TCP for Wireless Networks6. Equation-Based TCP Window Adjustment7. TCP Performance for Ad Hoc Networks8. Ad Hoc IP Routing over Bluetooth Scatternets9. Scalable DNS for Mobile Computing10. Active Ad Hoc Networking11. Energy-Aware Routing in Ad Hoc Networks12. Transparent Migration of Mobile Agents13. Service Discovery in Wireless Networks
A CS395T Course Project MobileComputing &
WirelessNetworks
Spring 2001C
ourse
we
b site
: http://w
ww
.cs.ute
xas.e
du
/use
rs/ygz/39
5T
• Instructo
r: Dr. Y
on
gg
uan
g Z
ha
ng
(ygz@
cs.ute
xas.ed
u)
Project Goals Implement & deploy of Mobile IP in CS
Network. Determining best location for home agent
and foreign agent. Study best mode of operation for mobile
node (co-located COA or FA assigned COA) to achieve smooth handoff.
Auto configuration of mobile node when network change is detected.
New Ideas Achieving smooth handoff while roaming
between different 802.11 wireless LANs operated in infrastructure point mode using mobile IP.
Algorithm to determine change of network–Guessing the network change based on three
missed beacon (registration request).–Guessing network change based on signal
strength–Based on users input
Mobile IP service for ACES and TAY
Results/Lesson Learned Experimental Mobile IP is operational in
CS Network Our Algorithm helped to achieve semi-
smooth handoff while moving between ACES and TAY wireless network (30 sec approx).
Lesson learned:–802.11 does not have good support for mobile
IP smooth handoff in infrastructure mode.–No easy heuristics to determine network
change (layer2) in 802.11 access point mode.
Project team: Madhusudan Kayastha, John ThomasURL: http://www.cs.utexas.edu/users/john/mobileIP.html
A CS395T Course Project MobileComputing &
WirelessNetworks
Spring 2001C
ourse
we
b site
: http://w
ww
.cs.ute
xas.e
du
/use
rs/ygz/39
5T
• Instructo
r: Dr. Y
on
gg
uan
g Z
ha
ng
(ygz@
cs.ute
xas.ed
u)
Key Design/Approach Data collation: mapping the service area in
advance for an accurate signal map Algorithm 1: best matching: O(n2) Algorithm 2: multidimensional search
O(log(n)) GUI (see above) Limitation : Placement of Base stations
limited to certain locations
New Ideas Increasing proliferation of 802.11 base
stations and PC card use 802.11 wireless LAN infrastructure to determine mobile device location inside building
Signal strength can be measured and there is a correlation between signal strength and distance
Signal strengths from multiple base stations can triangulate the exact location
Location Service for Mobile Device
Results/Lesson Learned Implemented in Linux Tested in ACES 6th floor
–Beacon stations: 2–Service granularity: 10 feet
Accuracy: 90% of the time: within 15 feet Lesson learned:
–802.11 Linux driver does not support “promiscuous” mode to pick up beacons from multiple base stations
–Resolved to ad-hoc mode
Project team: Ravishankar Chamarajnagar, Jeff NapperURL: http://www.cs.utexas.edu/users/ravshank/mobile_project/project.html
A CS395T Course Project MobileComputing &
WirelessNetworks
Spring 2001C
ourse
we
b site
: http://w
ww
.cs.ute
xas.e
du
/use
rs/ygz/39
5T
• Instructo
r: Dr. Y
on
gg
uan
g Z
ha
ng
(ygz@
cs.ute
xas.ed
u)
Design Generic Interface for all ad-hoc routing
protocols– Network Interface Module (NIM)– User Interface Module (UIM)
NIM: Interface to Network Stack AKM: In Data Plane
– Forwarding according to Ad-hoc protocol UIM: for data and control plane comm’n. AUD: In Control Plane
– Ad-hoc Route maintenance
New Ideas Many Ad-hoc Routing Protocols Exist Very few real implementations
– High Complexity– Not enough tools for development– Performance claims based on simulation
results Designed and developed a generic
infrastructure– Provides a tool to evaluate Ad-hoc protocols
Implementation & Testing Implemented in Linux 2.4.3
– Used Netfilter & IPTable support Tested multihop routing
–Three hosts in broadcast setup Test Protocols:
– Manual Configuration – AODV
Can be extended to support multiple protocols simultaneously
Project team: Puneet Chopra, Sumit Garg, Sugat JainURL: http://www.cs.utexas.edu/users/puneet/mobcomp/proposal.html
AKM:Ad-hoc Kernel ModuleAUD: Ad-hoc User DaemonART: Ad-hoc Routing Table
NIM: Network Interface ModuleUIM: User Interface Module
Linux Kernel Support for Ad-hoc Routing
A CS395T Course Project MobileComputing &
WirelessNetworks
Spring 2001C
ourse
we
b site
: http://w
ww
.cs.ute
xas.e
du
/use
rs/ygz/39
5T
• Instructo
r: Dr. Y
on
gg
uan
g Z
ha
ng
(ygz@
cs.ute
xas.ed
u)
Design Function separation between user and
kernel; Necessary communication overload
between user and kernel are reduced to minimum;
New Ideas Provide an architecture for ad hoc routing
protocol implementation by extracting the common part, like multihop for packet delivery, reliable link layer maintenance, etc into general kernel module.
Implementation & Testing Kernel on top of Linux 2.4.2; Two hosts can talk to each other; Communication between user and kernel
are tested; AODV user daemon implemented Other protocol implementation are
extendable
Project team: Zhenxun Xiao, Wei Li, Minesh ShilotriURL: http://www.cs.utexas.edu/users/liwei/AdHoc/
Linux Kernel Module for Ad-hoc Routing
A CS395T Course Project MobileComputing &
WirelessNetworks
Spring 2001C
ourse
we
b site
: http://w
ww
.cs.ute
xas.e
du
/use
rs/ygz/39
5T
• Instructo
r: Dr. Y
on
gg
uan
g Z
ha
ng
(ygz@
cs.ute
xas.ed
u)
Key Design/Approach A mode includes cwnd, ssthresh, rtt, srtt,
rttvar, rtxcur, etc. Generally, save the current mode and load
a new mode for a vertical handoff Schedule a handoff timer for a downward
handoff to avoid fast retransmission caused by out-of-order delivery
Reset mode for an upward handoff to an unvisited overlay network
An extension to New-reno TCP
New Ideas TCP manages multiple modes A mode is a subset of entire TCP state
– corresponds to an overlay network– dynamic
TCP is notified of the handoff events TCP switches between modes when the
mobile host vertical handoffs between different overlay networks
Different actions for upward and downward handoffs
Multi-Mode TCP for Wireless Networks
Results/Lesson Learned Implemented in ns2 Simulation results in ns2
–Eliminate unnecessary retransmissions caused by timeouts or duplicate acks
–TCP goodput improved (> 7%) Lesson learned:
– Implementation of TCP stack– Limitation of the simulation– Need of more realistic experiments
Project team: Fengfeng Tu, Hongxia TianURL: http://www.cs.utexas.edu/users/ftu//project.html
Motivation TCP has been tuned to traditional
networks comprising wired links and stationary hosts
Wireless overlay networks– Heterogeneous overlays– User mobility: vertical handoffs– Sudden, frequent, and significant changes in
bandwidth and delay Study TCP behavior over wireless overlay
networks and improve its performance
A CS395T Course Project MobileComputing &
WirelessNetworks
Spring 2001C
ourse
we
b site
: http://w
ww
.cs.ute
xas.e
du
/use
rs/ygz/39
5T
• Instructo
r: Dr. Y
on
gg
uan
g Z
ha
ng
(ygz@
cs.ute
xas.ed
u)
Design Objective: fast estimation of congestion
window size–Node i keeps its own congestion loss rate pi
–TCP packet has an additional field, FP (Forwarding Probability).
–FP is initially set to 1 , and multiplied by 1-pi at each intermediate node i.
–Receiver reports FP to the sender with ACK.–Congestion window is calculated from the
path loss rate p = 1-FP:
Problem Packet corruption caused by wireless
transmission–Caching and retransmission at base station.–Each node in ad hoc network doesn’t have
enough resource. Link failure due to mobility
–Explicit notification — freeze TCP states until new route found.
–Old states may not be appropriate for new route.
Results 2
Project team: Min Sik Kim, Eunjin JungURL: http://www.cs.utexas.edu/users/minskim/mobile/
Equation-Based TCP Window Adjustment
3p/2/1cwnd
Results 1
A CS395T Course Project MobileComputing &
WirelessNetworks
Spring 2001C
ourse
we
b site
: http://w
ww
.cs.ute
xas.e
du
/use
rs/ygz/39
5T
• Instructo
r: Dr. Y
on
gg
uan
g Z
ha
ng
(ygz@
cs.ute
xas.ed
u)
Simulation Setup Using NS2
–20 mobile nodes–500x500 grid–6 different levels of moving speed–Single and multiple connections traffic pattern
Problem What is the performance of TCP over an
ad hoc network, under each of the four ad hoc network routing protocols?
–DSR–AODV–DSDV–TORA
Project team: Feng Wang, Zhiwei LinURL: http://www.cs.utexas.edu/users/wangf/manetcp.htm
TCP Performance for Ad Hoc Networks
Results Conclusion DSR and DSDV is the best AODV is not as good TORA is far worst
A CS395T Course Project MobileComputing &
WirelessNetworks
Spring 2001C
ourse
we
b site
: http://w
ww
.cs.ute
xas.e
du
/use
rs/ygz/39
5T
• Instructo
r: Dr. Y
on
gg
uan
g Z
ha
ng
(ygz@
cs.ute
xas.ed
u)
Key Design/Approach IP networking for bluetooth scatternet
–Piconet bridging scheme–Each scatternet as one IP network–Host based addressing
Master Relay Switch Routing (MRSR)–Ad-hoc routing in scatternet–based on AODV
New Ideas A new architecture for IP networking in
bluetooth scatternet–Current standard (LAN profile, BNEP) does
not tackle inter-piconet IP networking issue Cluster-based on-demand routing strategy
–Master Relay Switch Routing (MRSR) protocol for inter-piconet IP forwarding
Ad Hoc IP Routing over Bluetooth Scatternets
Results/Lesson Learned Implemented in NS2 Extension to IBM BlueHoc
Project team: Sidharth Choudhury, Anupam RastogiURL: http://www.cs.utexas.edu/users/sidharth/report
A CS395T Course Project MobileComputing &
WirelessNetworks
Spring 2001C
ourse
we
b site
: http://w
ww
.cs.ute
xas.e
du
/use
rs/ygz/39
5T
• Instructo
r: Dr. Y
on
gg
uan
g Z
ha
ng
(ygz@
cs.ute
xas.ed
u)
Key Design/Approach
DNS Workload Analysis
4.5% in number, 0.8% in bandwidth
80% absorbed at client cache 70% absorbed at name servers
94% absorbed by cache–Problem if no cache
Scalable DNS for Mobile Computing
Results/Lesson Learned
Importance of smart caching scheme Scalability of DNS can be achieved from
server replication. Update individual hosts instead of zones
Project team: Dong-Young Lee , Bong-Soo SohnURL: http://www.cs.utexas.edu/users/bongbong/mobile
Motivation Using DNS as location service
infrastructure for mobile computing
Dynamic updates are already defined in RFCs
No caching? Scalability?
A CS395T Course Project MobileComputing &
WirelessNetworks
Spring 2001C
ourse
we
b site
: http://w
ww
.cs.ute
xas.e
du
/use
rs/ygz/39
5T
• Instructo
r: Dr. Y
on
gg
uan
g Z
ha
ng
(ygz@
cs.ute
xas.ed
u)
Key Design / Approach DSR(basic, optimized, multipath), AODV,
and TORA protocols implemented–Node Resident Service functions for basic
protocol operations (coded in Popcorn)–Protocols (really mobile glue code)
implemented in PLAN Dynamic Linking of protocol modules Limitation:
–Due to lack of concurrency in FASTNet, only Route Discovery protocol is implemented. Route Maintenance is emulated.
New Ideas
Dynamic Ad Hoc Routing Protocols implemented in an active networking test-bed
Dynamic Protocol Dissemination using mobile code
Provides Increased adaptability to unpredictable Ad Hoc environment
Active Ad Hoc Networking
Results / Lessons Learned Implemented in FASTNet & PLAN (Linux) Choose routing protocols adaptively. Lessons learned:
–AN can provide dynamic & efficient routing protocols for MANET
–Decision-making mechanism for optimal protocol in each environment is needed
–Need to extend this idea in test-bed system
Project team: Seong-Kyu Song, Stephen ShannonURL: http://www.ece.utexas.edu/~shannon/CS395T.htm
PLAN
PLAN
DSR AODV
TORA
Active (FastNet) Node
DSR
TORA
AODV
TORA
Code Pushing
Remote Evaluation
DSR
AODV
Active (FastNet) Node
Active (FastNet) Node
A CS395T Course Project MobileComputing &
WirelessNetworks
Spring 2001C
ourse
we
b site
: http://w
ww
.cs.ute
xas.e
du
/use
rs/ygz/39
5T
• Instructo
r: Dr. Y
on
gg
uan
g Z
ha
ng
(ygz@
cs.ute
xas.ed
u)
Simulation Setup
Single source-destination pair without node mobility
Single source-destination pair with node mobility Multiple source-destination pairs without node
mobility Multiple source-destination pairs with node
mobility Introduction of energy balance factor which
represents the heterogeneity of initial energy level at each node
Routing Algorithms
Maxmin routing algorithm – choose the path whose minimum energy level node has the maximum value among all paths.
Maxmax routing algorithm – choose the path which has maximal residual energy on the path
Maxavg routing algorithm – choose the path which has maximal average energy level
Shortest path routing algorithm – no energy consideration
Energy-Aware Routing in Ad Hoc Networks
Results/Lesson Learned Overall our Maxmin algorithm performs well. One example of results:
Project team: Jangwon Lee & Xiangying YangURL: http://www.ece.utexas.edu/~yangxy/395T.htm
Motivation Mobile nodes are battery-operated and one or
several nodes’ energy exhaustion may cause the disruption of the entire network
A number of approaches focus on minimizing the total consumed power.
Several approaches’ goal was to maximize the life time of the system. However, these are based on static topology and given traffic demands in advance.
Our goal is to propose energy-aware routing algorithms in dynamic ad hoc network environments to achieve max. system lifetime.
Simulation techniques such as generating topology and random traffic etc.
Obtain a research trend of energy-aware design in mobile computing environments
A CS395T Course Project MobileComputing &
WirelessNetworks
Spring 2001C
ourse
we
b site
: http://w
ww
.cs.ute
xas.e
du
/use
rs/ygz/39
5T
• Instructo
r: Dr. Y
on
gg
uan
g Z
ha
ng
(ygz@
cs.ute
xas.ed
u)
Design/Approach Provide a Java mobile agent system that
supports transparent migration Use standard JVM to gain better
acceptance Use byte-code instead of source code
New Ideas Transform mobile agent byte-code to
support transparent migration Simulate thread and stack frame to
support saving and restoring call stack and local variables
Mechanism for surviving migration failure which provides more reliable disconnected operation
Transparent Migration of Mobile Agents
Results/Lesson Learned Developed API, parser, and server for
mobile agent programming and deployment
With transparent migration, mobile agent programming is more intuitive
Demonstrate typical mobile agents applications using our framework
Project team: Endi S. Dewata, Kiran K. AdduriURL: http://www.cs.utexas.edu/users/endisd/mobile
agent source code
agent byte-code
agent byte-codewith supportfor mobility
javac
parser
server
agents
transparentmigrationwith RMI
A CS395T Course Project MobileComputing &
WirelessNetworks
Spring 2001C
ourse
we
b site
: http://w
ww
.cs.ute
xas.e
du
/use
rs/ygz/39
5T
• Instructo
r: Dr. Y
on
gg
uan
g Z
ha
ng
(ygz@
cs.ute
xas.ed
u)
Key Design / Approach Component-based technologies. Multimedia Services Application:
–Multimedia Client: Plays multimedia files from Service URLs using JMF.
–Multimedia Service Provider: Provides Service URLs to clients, interfaces to user database.
–Reggie (The Service Lookup by Sun Microsystems): Keeps track of available services.
–User Database (Microsoft Access connected to Service Provider via ODBC and JDBC)
New Ideas A Jini-based Multimedia Services
Application designed for ad hoc wireless systems.
Use of multicast-based service discovery protocols for multimedia services in ad hoc wireless networks.
–Adapt to changes in bandwidth in wireless overlay networks.
–Provide value-added services to roaming users with minimal configuration.
Integrated user accountability for services.
Service Discovery in Wireless Networks
Results / Lessons Learned Service discovery is crucial in emerging
packet-based wireless technologies (3G/4G, Bluetooth)
Lessons learned:–Holding users accountable for using wireless
bandwidth for value-added services is necessary.
–Key challenge: Integrating various enabling technologies (Jini, JMF, JDBC, etc.)
Project team: Dogu Arifler, Vivek SharmaURL: http://www.ece.utexas.edu/~arifler/CS395T/index.htm
Multimedia Provider
Java-based Client GUI
Microsoft Access
Lookup Service
Multicast in a Jini-enabled wireless network
SQL, JDBC
Service URL
Contact
Prof. Yongguang ZhangDept. of Computer SciencesUniversity of Texas at AustinAustin, Texas 78712, U.S.A.
E-mail: [email protected]: 512-232-7889