MobileMAN meeting Helsinki 7.6.2004

1
MobileMAN meetingHelsinki 7.6.2004
Jose Costa-Requena, Nicklas Beijar, Raimo
Kantola Jose, Nicklas.Beijar, Raimo.Kantola_at_netl
ab.hut.fi Networking Laboratory Helsinki
University of Technology P.O. Box 3000,
FIN-02015, Finland
2
Preliminary architecture Main modules

• Context Roaming Module
• Provides the API to applications to roam
  depending on context information
• Ad Hoc Framework
• Implements the different ad hoc routing
  algorithms (proactive, reactive, or others)
• Interacts with the Linux Kernel for implementing
  low level routing procedures.
• Supports necessary extensions to perform service
  discovery or routing optimisations.
• Consists of one Common module and several Routing
  modules

Terminal Applications
Context Roaming
Module
AdHoc_Framework
Kernel
Interface
Handler
3
Architecture
Completed
Initiated not completed (work in progress)
Terminal Applications
Dropped
VoIP (SIP) Real Time QoS (RTCP)
AdHoc_Framework
Routing Module
Service Module
Routing Module
ZRP
MobileMAN
Spearhead(IRP)
Common Module
Routing Module
Routing Module
Service Module
AODV
OLSR
Common Cache
RoutingService Data
Ad Hoc Framework API
MAC watchdog
Generic Ad Hoc Module
Common Module API
Access kernel routes Retrieve MAC info
Kernel Ad Hoc API
4
ZRP

• The ZRP is being developed on top of AODV and
  OLSR.
• Initial testing but no results available yet.
• The Ad Hoc framework allows updating each
  protocol without affecting the overall
  functioning.
• New version of AODV (v0.8) has been updated into
  the framework (better performance).

ZRP
IARP
IARP
IERP
IERP
Logic
Logic
Logic
Logic
Protocol Specific
Protocol Specific
Protocol Specific
Protocol Specific
Adapter
Adapter
Adapter
Adapter
Common
Cache

Registry
Service
CCRS
Proxy
CCRS
Proxy
OLSR
BRP
AODV
5
Service Discovery

• Reactive service discovery mechanism implemented
  but not integrated yet.
• Implemented the node classification for the nodes
  that will create the service discovery backbone.
• Smart-dummy node architecture for service
  discovery infrastructure implemented but not full
  functional yet.
• Missing functionality
• Establish link state using OLSR
• Define service information as part of OLSR link
  state data
• Manage AODV service queries.

6
Node classification

• During this period we have implemented the node
  classification (i.e. smart, dummy nodes) for
  creating Ad Hoc backbone to support other
  constrained devices and extend the network
  lifetime.
• Classification criteria is based on
• Link stability, Mobility, Battery power and User
  preferences
• This node classification should be flexible and
  dynamic in order to allow nodes to change roles
  (e.g. 'Smart' node becomes a 'Dummy' node)
• The Attach Detach procedure implemented with
  AODV.
• The smart nodes will establish link state
  relationships using OLSR and will share routing
  and service information.

7
Applications

• The main task during last period was to test
  voice sessions in the Ad Hoc framework.
• 1) Ready made stack Vovida SIP client ported to
  iPaq
• Works with iPaq 3950 (model)
• Doesnt work with iPaq 3850 (model) due to
  drivers problems
• Ad hoc routing currently tested with only one-hop
  because lack of equipment.
• The stack requires 5Mb plus excessive CPU
  resources.
• 2) Light-weight SIP stack under development
• Binary around 1Mb versus 5Mb of ready made Vovida
  stack
• Possibility to modify the media transport to
  improve performance in multi-hop Ad Hoc networks.

8
VoIP session testing

• Test case 1
• Test-bed includes two nodes in a direct
  communication between them and a third node with
  Ethereal acting as sniffer to gather all the
  measurement data.
• Metrics Use of timing information and a sequence
  number in RTP header.
• packet delay
• Packets loss
• Sequence error
• Jitter

9
Test case 1
0,0203

• Nodes separated 2m, 10m and 40m
• Results
• Same mean delay (20ms.)
• Jitter constant value between 29ms. and 34ms.
• Packet loss less than 1 (0.822).

0,02028231
0,02028
0,02026
0,02024
0,020222211
0,02022
0,020208559
Time (sec)
0,0202
0,02018
0,02016
1
2
3
Test
10
Test case 2

• Including obstacles in the test environment.

11
Test case 2

• Results compared to Test 1
• Similar mean delay (20.243ms).
• Similar packet loss.
• Similar jitter value between 12ms. and 34ms.
• Test 1
  Test 2

12
Test Case 3

• Test case 3 has been performed but data has not
  been compiled yet.

13
Conclusions

• The results show that media stream flow smoothly
  independently of the distance and obstacles.
• However, the user perception is bad due to lack
  of processing resources in the nodes, which delay
  the media processing.
• To be verified
• The processing delay increases when node
  resources decrease.
• The delay also increases in multi hop routes.
• To be added
• Light stack for real time (RT) applications
  optimised for low resources nodes.
• QoS module that signals increases in packet delay
  and triggers optimal route discovery.
• Establish and discover routes with nodes that
  have enough resources for RT sessions.

14
Summary current status

• AODV routing protocol (updated)
• Completed
• OLSR routing protocol
• Completed
• ZRP routing protocol
• Initial testing, bug fixing
• Node classification
• Designed, algorithm developed
• Vovida SIP stack
• Test results
• Own SIP stack
• Code authoring

• Service discovery
• Reactive discovery implemented.
• Proactive service distribution (work under
  progress).
• Virtual backbones
• Under research
• Simulations
• Halted
• MAC layer driver
• Halted
• Location assisted routing
• Halted

15
Further work

• Focus on
• Finalise existing routing modules (i.e. ZRP)
• Complete the light-weight SIP stack under
  development with QoS optimisation.
• Create an Ad Hoc network based on groups of smart
  nodes that will create an ad hoc backbone
• Finalise the service discovery module (reactive
  service queries from dummy nodes and proactive
  service information exchange between smart
  nodes).
• Implement and integrate the cross-layer Service
  Discovery mechanism.

16
Questions?

• Thank you for
• your attention!

17
Task force
Full time (4)
Part time (5)
Work completed (2)
MobileMAN
HUT (Raimo Kantola, Jose Costa-Requena, Nicklas
Beijar)