[Trac #1014] Low throughput in mesh mode

[Zarro Boogs per Child]

#1014: Low throughput in mesh mode
----------------------+-----------------------------------------------------
 Reporter:  rchokshi  |        Owner:  rchokshi 
     Type:  defect    |       Status:  new      
 Priority:  normal    |    Milestone:  Untriaged
Component:  wireless  |   Resolution:           
 Keywords:            |  
----------------------+-----------------------------------------------------
Changes (by jcardona):

  * component:  distro => wireless

Comment:

 Throughput measurements should be averaged out.  Some measurements taken
 in our office:


 {{{
 mesh:[  3]  0.0-10.0 sec  8.10 MBytes  6.77 Mbits/sec
 ad-hoc:[  3]  0.0-10.0 sec  9.53 MBytes  7.97 Mbits/sec
 mesh:[  3]  0.0-10.1 sec  7.42 MBytes  6.19 Mbits/sec
 ad-hoc:[  3]  0.0-10.0 sec  14.8 MBytes  12.4 Mbits/sec
 mesh:[  3]  0.0-10.2 sec  1.64 MBytes  1.35 Mbits/sec
 ad-hoc:[  3]  0.0-10.0 sec  4.88 MBytes  4.09 Mbits/sec
 mesh:[  3]  0.0-10.1 sec  1.02 MBytes    849 Kbits/sec
 ad-hoc:[  3]  0.0-10.0 sec  6.42 MBytes  5.37 Mbits/sec
 mesh:[  3]  0.0-10.0 sec  7.73 MBytes  6.48 Mbits/sec
 ad-hoc:[  3]  0.0-10.0 sec  14.8 MBytes  12.4 Mbits/sec
 mesh:[  3]  0.0-10.3 sec  1.15 MBytes    938 Kbits/sec
 ad-hoc:[  3]  0.0-10.0 sec  5.37 MBytes  4.49 Mbits/sec
 mesh:[  3]  0.0-10.0 sec  6.96 MBytes  5.82 Mbits/sec
 ad-hoc:[  3]  0.0-10.0 sec  12.9 MBytes  10.8 Mbits/sec
 mesh:[  3]  0.0-10.0 sec  8.40 MBytes  7.03 Mbits/sec
 ad-hoc:[  3]  0.0-10.0 sec  13.0 MBytes  10.9 Mbits/sec
 mesh:[  3]  0.0-10.0 sec  8.13 MBytes  6.79 Mbits/sec
 ad-hoc:[  3]  0.0-10.0 sec  11.3 MBytes  9.43 Mbits/sec
 mesh:[  3]  0.0-10.0 sec  7.27 MBytes  6.07 Mbits/sec
 }}}

 Or more readable:
 {{{
 Mesh:
  7.03 Mbits/sec
  6.77 Mbits/sec
  6.79 Mbits/sec
  6.48 Mbits/sec
  6.19 Mbits/sec
  6.07 Mbits/sec
  5.82 Mbits/sec
  1.35 Mbits/sec
    849 Kbits/sec
    938 Kbits/sec

 Ad-hoc:
  12.4 Mbits/sec
  12.4 Mbits/sec
  10.9 Mbits/sec
  10.8 Mbits/sec
  9.43 Mbits/sec
  7.97 Mbits/sec
  5.37 Mbits/sec
  4.49 Mbits/sec
  4.09 Mbits/sec
 }}}

 The script used for the test:

 {{{
 while true
 do
         echo -n mesh:
         iperf -c 192.168.3.77 | grep '/sec'
         echo -n ad-hoc:
         iperf -c 192.168.2.77 | grep '/sec'
 done
 }}}


 To see that the right interface is used for each address:
 {{{
 # ip addr show to 192.168.3.1/24
 6: msh0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast qlen
 1000
     inet 192.168.3.78/24 brd 192.168.3.255 scope global msh0

 -bash-3.1# ip addr show to 192.168.2.1/24
 5: eth1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast qlen
 1000
     inet 192.168.2.76/24 brd 192.168.2.255 scope global eth1
 }}}

 So, in summary, the maximum throughput that we can get from the mesh at
 this time is 45% slower than the maximum measured ad-hoc rate.  What we
 should determined is the cause for the minimum observed rates.  Suspecting
 that those values occur when the route is being refreshed.  To confirm, we
 should correlate with sniffer captures.

-- 
Ticket URL: <http://dev.laptop.org/ticket/1014#comment:5>
One Laptop Per Child <http://laptop.org/>