MathiasKurthRestricted

Hardware

• AMD Geode

Software

• Kernel Stuff
• SeisComP (ftp, docu)

Simulation

• Java in Simulation Time

Early Warning

• Ideas on EWS

Devilish Opportunistic Routing

• Beobachtung: Kein Candidate Set ist eine große Einschränkung des Opportunismus.
• Idee: Das Candidate Set wird nicht vorgegeben, sondern 'dynamisch' ermittelt.

Dadurch entstehen die folgenden Probleme:

• Richtung: Das Paket muss mit jedem Hop einen Fortschritt richtung Final Destination machen.
• Ack Ordering: Die Menge der Knoten, die das Paket empfangen haben, muss sich auf eine kollisionsarme Ack-Reihenfolge einigen.
• Größe das Candidate Set: Potentiell sind jetzt alle Nachbarn Kandidaten. Das Ack-Fenster muss aber beschränkt bleiben (Das Ack hat feste Größe und feste Übertragungszeit.)

Ein Knoten s sendet folgendes Frame:

|------------------------------------|
| Final Destination d                |
|------------------------------------|
| ETX(d)_max | ETX(d)_min            |
|------------------------------------|
| Source Address s                   |
|------------------------------------|
| ...                                |
| Aorte Route                        |
| ....                               |              
|------------------------------------|

d - Address of the final Destination
s - Address of the source (sender)
ETX(d)_max - cuml. ETX from s to d
ETX(d)_min - Min(cuml. ETX from Neighbor(s) to d)
Aorta Route - Best route from s to d (hop-count) with ETX values

Resultate der Messungen im internen Testbed

• Measurement Results
• Auswertung der Delivery Ratio Messung
• Performance of a DHT Implementation in the BRN Indoor Testbed

Todo

• JiST
  • Multichannel-Mac
  • Javis & EventGUI
  • Measured, predefined Link Quality in Fields (Ersetzen des Ausbreitungsmodells gegen unsere Meßdaten)
  • How to measure with JiST/SWANS (generate diagrams & co.)
  • ExOR: slotted ack
    • ns2:
  • (Network Coding)

• Administrations-Tools für das Indoor-Testbed
  • Long term monitoring (cacti, rrdtool)
• Telnet-Zugang auf Knoten des Indoor-Testbed ohne Ethernet
• Methode zur Durchführung von Tests im Indoor-Testbed
  • z.B. [1]

Each experiment measures throughput between 65 randomly
selected node pairs. First, the nodes broadcast 1500-
byte packets every ten seconds for ten minutes and report
the measured delivery probabilities from all other nodes to
a central server. The server distributes this information
to all the nodes. The measurements are used to compute
ETX metrics and traditional routes. Next, the server contacts
each of the 65 node pairs in sequence, telling the pair
to measure the time required to transfer a 1.0 megabyte
¯le using traditional routing, then to wait 15 seconds, then
to measure the time required to transfer 1.1 megabytes using
ExOR. The evaluation does not use the combination of
ExOR and traditional routing, so the extra 0.1 megabyte is
to compensate for the 10% of packets which may not have
been delivered ordinarily. The reported throughput is one
megabyte divided by the total time required to transfer the
data. Every twenty minutes, the central server suspends the
experimental runs to recollect the link loss measurements.
During the experiment, existing Roofnet routing and user
tra±c are present.