Modify

Opened 4 years ago

Closed 3 years ago

#16997 closed defect (no_response)

Throughput Drop in Later Revisions of OpenWrt on TL-WR842ND V1

Reported by: furryfur1@… Owned by: developers
Priority: high Milestone: Chaos Calmer 15.05
Component: kernel Version: Trunk
Keywords: Cc:

Description

This issue is present in both Barrier Breaker trunk and the latest revisions of AA built from the git repository. The throughput drop occurs when going from r37917 to r37918 or any later revision in the BB trunk, r37917 does not exhibit the throughput drop but all later revisions do. The throughput drop also occurs when going from AA revision r39927 to r39928 or later. The throughput drop is about 10Mbps in my test-bed: between two routers, the transmitting router in WDS mode, both running OpenWrt, the main transmit router is a TL-WR842ND V1 and is the only one I changed firmware on to try to narrow down which revisions were causing the drop. I patched each revision before building it with my 40MHz WDS patch (stop wpa_supplicant locking hostapd to 20MHz, #12720) so that the test was fair against later revisions which have this patch, as it affects throughput. I also tested throughput between the main transmit router and other devices (Android tablet running iperf, 20MHz only, not compatible with my upper channel) and noted I also got reduced throughput here, although not as much as the devices were a lot closer. I am using reghack to unlock channel 13 which should be available in my country on each build.

The iperf command from the main transmit router was as follows: iperf -w 320k -t 30 -c 192.168.2.3. The iperf command on the receiving router was as follows: iperf -w 320k -s. An example of the throughput difference between revisions is as follows: Transmitter (trunk r41181) ---> Receiver (AA r36088) : 32.6Mbps; Transmitter (AA r36088) ---> Receiver (AA r36088) : 43.5Mbps. I did not record all my experiments however they were consistent on affected revisions, dropping about 10Mbps.

This router is also connected via WDS to another upstream router which I used to establish that changing revisions did not affect receive throughput, only transmitter throughput (throughput from upstream router seems to stay the same no matter what revision I use on the downstream). I notice an overall reduced throughput in my internet throughput which comes from the upstream router as well (upstream running BB r41181, also on a TL-WR842ND V1).

/etc/config/wireless for main transmit router
config wifi-device 'radio0'

option type 'mac80211'
option macaddr '90:f6:52:78:19:4c'
option hwmode '11ng'
list ht_capab 'SHORT-GI-20'
list ht_capab 'SHORT-GI-40'
list ht_capab 'TX-STBC'
list ht_capab 'RX-STBC1'
list ht_capab 'DSSS_CCK-40'
option rts '500'
option channel '9'
option htmode 'HT40+'
option noscan '1'
option log_level '1'
option distance '185'
option txpower '21'
option country '00'

config wifi-iface

option ssid 'UKD'
option device 'radio0'
option mode 'sta'
option key '#######'
option wds '1'
option encryption 'psk2+ccmp'
option network 'lan'
option bssid 'A2:F3:C1:9C:F8:76'

config wifi-iface

option device 'radio0'
option mode 'ap'
option ssid 'UKD'
option wds '1'
option network 'lan'
option encryption 'psk2+ccmp'
option key '#######'

Attachments (0)

Change History (5)

comment:1 Changed 4 years ago by nbd

please try if adding this patch resolves the throughput regression: http://nbd.name/950-rts_test.patch

comment:2 Changed 4 years ago by furryfur1@…

It appears that my experiments were not controlled enough as the throughput from the remote upstream router to my downstream router (both of these are TL-WR842ND V1 routers) is not improved by flashing both to AA r39927. I flashed both the upstream router and the downstream from AA r39927 to AA r36088 with my 40MHz patch applied and the speed again increased from 21.1 Mbps to 33.3 Mbps, tested using iperf as above. It is very difficult to test the remote router (which would be ideal) with different versions of OpenWrt as it is hard to get it to maintain connection after flashing, not to mention it sometimes bricks when flashing remotely and it is mounted up a pole so it's difficult to access. I may borrow one of my neighbours routers in order to test the throughput between two TL-WR842ND V1s on different OpenWrt revisions so that the test configuration is identical to the configuration between my upstream and downstream routers. I have not tried the BB versions I said didn't exhibit the issue on the remote upstream router yet. I will test this patch to see if it fixes things once I get another TL-WR842ND for testing.

comment:3 Changed 4 years ago by nbd

any update on testing this?

comment:4 Changed 4 years ago by nbd

ping?

comment:5 Changed 3 years ago by nbd

  • Resolution set to no_response
  • Status changed from new to closed

Add Comment

Modify Ticket

Action
as closed .
The resolution will be deleted. Next status will be 'reopened'.
Author


E-mail address and user name can be saved in the Preferences.

 
Note: See TracTickets for help on using tickets.