#11133 NORM 1.75-so: XO-1.75 test wireless performance

[Zarro Boogs per Child]

#11133: XO-1.75 test wireless performance
-------------------------------------+--------------------------------------
           Reporter:  Quozl          |       Owner:                                   
               Type:  task           |      Status:  new                              
           Priority:  normal         |   Milestone:  1.75-software                    
          Component:  not assigned   |     Version:  Development build as of this date
         Resolution:                 |    Keywords:                                   
        Next_action:  test in build  |    Verified:  0                                
Deployment_affected:                 |   Blockedby:                                   
           Blocking:                 |  
-------------------------------------+--------------------------------------

Comment(by greenfeld):

 If there is a regression concern I need more information on how to
 reproduce it and how the laptops were setup/spaced.

 Local setup:
  * Two XO-1.75 B1's, 11.3.1 os26
  * One XO-1.5, 11.3.0 os883
  * WNDR3700 AP w/ wired computer, running Fedora 14+iperf on a semi-decent
 system.  The 2.4 GHz AP channel was limited to 54 Mbps because one (non-
 test) client I have cannot see the AP if set higher, but shouldn't matter
 since XO's do not support 802.11n at this time.
  * Systems kept ~1 meter from the AP or other XOs (for Ad-hoc) for testing

 Local (non-office) testing with the above setup using the wired computer &
 AP for an iperf server shows sampled throughput values ranging from ~14-16
 Mbits/sec when using an XO-1.75 B1 with WPA2-PSK active, and ~15-17
 Mbits/second using an XO-1.5.  Disabling encryption did not noticeably
 change the XO-1.5 results but did increase the XO-1.75's results to be
 more in line with the XO-1.5.

 Switching the XO-1.5 & 1.75 to talk to each other over ad-hoc using the
 XO-1.75 as the iperf server resulted in tests showing ~13.5-15.0
 Mbits/sec.  I then tried two XO-1.75 B1's talking to each other on all
 three Sugar ad-hoc network channels (1/6/11) with similar results.

 I did not test two XOs talking through the AP as I do not know if that is
 a concern with this ticket, and collision handling/backoff may vary from
 AP to AP.  I also do not know if my results are the expected limit,
 although they seem higher than everyone else's in this ticket so far.

 So there might be a ~1 MB/sec performance loss from XO-1.5 to 1.75 given a
 short distance between devices, but to prove that needs more
 analysis/samples/knowledge of what other 2.4 GHz users in the area are
 doing/etc.

-- 
Ticket URL: <http://dev.laptop.org/ticket/11133#comment:12>
One Laptop Per Child <http://laptop.org/>
OLPC bug tracking system