Wireless Sparse Testbed: Difference between revisions

From OLPC
Jump to navigation Jump to search
No edit summary
Line 16: Line 16:
=== Testbed Journal ===
=== Testbed Journal ===
Oct, 16 - More nodes are being added to connect the cloud, currently there are 3 disconnected sectors
Oct, 16 - More nodes are being added to connect the cloud, currently there are 3 disconnected sectors
Nov 1-15 - Two more nodes were added to the testbed in order to connect the disconnected sectors. Now, it is possible to ping a node from any other node in the testbed. Tests were performed (throughput, latency, packet loss) between nodes 4C and 2A with poor results (losses ranging from 30% to 90%). The group started to investigate the reasons of the poor connectivity (interference, busy spectrum, route churnning, etc).

Revision as of 14:01, 19 November 2008

This page is dedicated to the TestBed installed at Universidade Federal Fluminense, Brazil.

Professor Luiz Magalhães and a group of students of Laboratório Midiacom, with the participation of Ricardo Carrano, from OLPC, built this testbed that is designed to study the behavior of a sparse mesh network, and answer the following questions:

  • Is a sparse network feasible? What are the practical distances and what can be expected in term of throughput and latency?
  • How does the shared Internet mechanisms (aka MPP) behave and what happens when you have multiple gateways?
  • What is the effect of mobility on a sparse network?
  • What is a good adaptive behavior to multicast/broadcast transmission rates?
  • What is the efficiency of the current multicast implementation?
  • How OLPC-Mesh compares to the current 802.11s implementation on the Linux Kernel?

The testbed is currently composed of 12 B2-1 prototypes, running the 8.2 distro in a non-gui runlevel.

Sparse testbed.jpg

Testbed Journal

Oct, 16 - More nodes are being added to connect the cloud, currently there are 3 disconnected sectors Nov 1-15 - Two more nodes were added to the testbed in order to connect the disconnected sectors. Now, it is possible to ping a node from any other node in the testbed. Tests were performed (throughput, latency, packet loss) between nodes 4C and 2A with poor results (losses ranging from 30% to 90%). The group started to investigate the reasons of the poor connectivity (interference, busy spectrum, route churnning, etc).