[Testing] Connectivity Test Discussion - test plan from Marvell
Ronak Chokshi
rchokshi at marvell.com
Thu Jun 7 21:51:59 EDT 2007
Hi Kim and team,
I have listed most of the test cases that we have been working on and
testing. Please feel to add / modify / delete the test cases based on
our discussions on the Wednesday conf. call.
http://wiki.laptop.org/go/802.11s_Connectivity_Test_Plan#Test_cases_from
_Marvell_for_the_OLPC_XOs
Thanks
Ronak
________________________________
From: Kim Quirk [mailto:kim.quirk at gmail.com]
Sent: Wednesday, June 06, 2007 5:38 PM
To: Ronak Chokshi; Rajesh Bhagwat; Raja Banerjea; Ramya Dharshini
Chandrasekaran; javier at cozybit.com
Cc: office at laptop.org; Alan Cunningham; Dan Goldish; Christopher
Blizzard; Dan Williams; Marcelo Tosatti; Richard Smith; Andres Salomon;
William Mitchell Bradley; John (J5) Palmieri; Marco Pesenti Gritti;
eben.eliason at gmail.com; testing at laptop.org; miguel at laptop.org
Subject: Re: Connectivity Test Discussion, Minutes from 6/6/07
Minutes (also posted on internal wiki, 'Rollout and Deployment')
June 6, 2007
Attending: Michail, Kim, Jim, Walter, Dan (rh), Miguel, Chris (rh),
Cameron (rh), Zack (rh), Rajab (mv), Ronak (mv), Raja (mv)
* Michail went through his low level 802.11s test plan:
<http://wiki.laptop.org/go/802.11s_Connectivity_Test_Plan
>
* Path Discovery
* Used ping, ran tests over several hours, introduced new
nodes
* Found some lost pings in a noisy RF environment; no loss
in anachoic chamber
* Network randomly creates mesh paths if you don't force
the paths
* Communications Integrity
* Frame size (large frames) caused problems, so you need
to use scp to ensure large frames
* Performance testing
* Used Iperf
* To establish a baseline need to use UDP between 2 nodes
using 'infrastructure mode'
* Found that XO performance does not go down as quickly as
1/n due to efficiencies in node-to-node connections
* Very difficult to make a repeatable non-forced mesh
topology
* Need to set up tests to look for starvation or memory
issues (over time)
* Walter suggested some usage cases:
* 60 laptops, everyone talking to everyone
* 60 laptops, everyone talking to a server
* 15 groups of 4, or 4 groups of 15 (representing kids
collaborating in smaller groups)
* Groups defined geographically (clusters)
* Non-cluster Groups (with remote participants)
* Consider classroom topologies:
* Rows of students
* Circle of students
* Clusters
* Tests to be added:
* Multicast and broadcast
* Non realtime
* Realtime
* Dynamic nature of the real world
* Nodes coming and going
* Reconfiguration of the network
* Action items:
* Marvell India team has 20 laptops; Kim help them to get
40 more B2-2
* India is building up knowledge and an automated test bed
* Marvell with add their test plan to the wiki
* Cozybit will provide programmable anycast address
* Items for Trial-2 schedule:
o
* Operating while suspended
* Mesh beacons
* Programmable anycast
* Ascii entry for WEP keys
* Finish support for WiFi compatibility (WPA)
IPv6 discussion
The applications driving IPv6 are:
* management of remote XOs
* allowing XOs to find other XOs around the world.
Need to understand what this requires for XO and for Server and
implement.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.laptop.org/private/testing/attachments/20070607/a38a40da/attachment-0001.htm
More information about the Testing
mailing list