FB 4040 help for SQM for gaming

Hi all , i'm pretty much new to openwrt , i flashed yesterday for the first time my fritz box 4040.
The primary purpose of this "upgrade" was to fix the bufferbloat while gaming, i saw an improvement but not as mush as i hoped.
I followed the documentation instruction and set up the sqm, cake with piece of cake even tried "test triple isoleted llt cake" with better result but i'm unsure about Link layer adaptation as my connection is a FTTW and it's not mentioned in the documentation.
And i also have huge upload speed spikes: my cap should be 4Mb but it goes up to 8Mb and right after below 1Mb (210ms of bufferbloat)... lower the Upload speed in Smq doesn't help at all. Is there a way to cap my upload speed below 4mb for good?
Thank you :smiley:

uci show sqm
tc -s qdisc show

( paste output using the code tag icon < / > )
( you can use luci-app-commands if ssh is new to you )

you could paste the result of a http://www.dslreports.com/speedtest ( or similar ) using the stock firmware / alternate device ... or even better a directly connected laptop if that's possible...

With luci-app-commands i got Command failed (Code: 65280), probably my fault i don't know how to do it.

Old dsl : http://www.dslreports.com/speedtest/64500887

Now : http://www.dslreports.com/speedtest/64792372

tc -s qdisc showsqm.@queue[0]=queue
sqm.@queue[0].debug_logging='0'
sqm.@queue[0].verbosity='5'
sqm.@queue[0].interface='eth1'
sqm.@queue[0].qdisc_advanced='0'
sqm.@queue[0].qdisc='cake'
sqm.@queue[0].enabled='1'
sqm.@queue[0].linklayer='none'
sqm.@queue[0].script='piece_of_cake.qos'
sqm.@queue[0].download='32000'
sqm.@queue[0].upload='3000'
root@OpenWrt:~# tc -s qdisc show
qdisc noqueue 0: dev lo root refcnt 2
 Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
 backlog 0b 0p requeues 0
qdisc mq 0: dev eth0 root
 Sent 17036246012 bytes 13223274 pkt (dropped 0, overlimits 0 requeues 0)
 backlog 0b 0p requeues 0
qdisc fq_codel 0: dev eth0 parent :4 limit 10240p flows 1024 quantum 1514 target 5.0ms interval 100.0ms memory_limit 4Mb ecn
 Sent 3963217 bytes 4655 pkt (dropped 0, overlimits 0 requeues 0)
 backlog 0b 0p requeues 0
  maxpacket 0 drop_overlimit 0 new_flow_count 0 ecn_mark 0
  new_flows_len 0 old_flows_len 0
qdisc fq_codel 0: dev eth0 parent :3 limit 10240p flows 1024 quantum 1514 target 5.0ms interval 100.0ms memory_limit 4Mb ecn
 Sent 5149212 bytes 15257 pkt (dropped 0, overlimits 0 requeues 0)
 backlog 0b 0p requeues 0
  maxpacket 0 drop_overlimit 0 new_flow_count 0 ecn_mark 0
  new_flows_len 0 old_flows_len 0
qdisc fq_codel 0: dev eth0 parent :2 limit 10240p flows 1024 quantum 1514 target 5.0ms interval 100.0ms memory_limit 4Mb ecn
 Sent 24246914 bytes 60002 pkt (dropped 0, overlimits 0 requeues 0)
 backlog 0b 0p requeues 0
  maxpacket 0 drop_overlimit 0 new_flow_count 0 ecn_mark 0
  new_flows_len 0 old_flows_len 0
qdisc fq_codel 0: dev eth0 parent :1 limit 10240p flows 1024 quantum 1514 target 5.0ms interval 100.0ms memory_limit 4Mb ecn
 Sent 17002886669 bytes 13143360 pkt (dropped 0, overlimits 0 requeues 0)
 backlog 0b 0p requeues 0
  maxpacket 4542 drop_overlimit 0 new_flow_count 2 ecn_mark 0
  new_flows_len 0 old_flows_len 1
qdisc cake 80a6: dev eth1 root refcnt 5 bandwidth 3Mbit besteffort triple-isolate nonat nowash no-ack-filter split-gso rtt 100.0ms raw overhead 0
 Sent 1246981 bytes 6314 pkt (dropped 5, overlimits 4398 requeues 0)
 backlog 330b 5p requeues 0
 memory used: 181760b of 4Mb
 capacity estimate: 3Mbit
 min/max network layer size:           42 /    1494
 min/max overhead-adjusted size:       42 /    1494
 average network hdr offset:           14

                  Tin 0
  thresh          3Mbit
  target          6.1ms
  interval      101.1ms
  pk_delay        3.9ms
  av_delay        889us
  sp_delay         33us
  backlog          330b
  pkts             6324
  bytes         1250533
  way_inds            0
  way_miss           27
  way_cols            0
  drops               5
  marks               0
  ack_drop            0
  sp_flows            5
  bk_flows            1
  un_flows            0
  max_len         10080
  quantum           300

qdisc ingress ffff: dev eth1 parent ffff:fff1 ----------------
 Sent 12568356 bytes 9413 pkt (dropped 0, overlimits 0 requeues 0)
 backlog 0b 0p requeues 0
qdisc noqueue 0: dev br-lan root refcnt 2
 Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
 backlog 0b 0p requeues 0
qdisc noqueue 0: dev wlan0 root refcnt 2
 Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
 backlog 0b 0p requeues 0
qdisc cake 80a7: dev ifb4eth1 root refcnt 2 bandwidth 32Mbit besteffort triple-isolate nonat wash no-ack-filter split-gso rtt 100.0ms raw overhead 0
 Sent 12720094 bytes 9390 pkt (dropped 8, overlimits 6540 requeues 0)
 backlog 19424b 15p requeues 0
 memory used: 121544b of 4Mb
 capacity estimate: 32Mbit
 min/max network layer size:           60 /    1494
 min/max overhead-adjusted size:       60 /    1494
 average network hdr offset:           14

                  Tin 0
  thresh         32Mbit
  target          5.0ms
  interval      100.0ms
  pk_delay       11.0ms
  av_delay        4.2ms
  sp_delay        578us
  backlog        19424b
  pkts             9413
  bytes        12751150
  way_inds            0
  way_miss           27
  way_cols            0
  drops               8
  marks               0
  ack_drop            0
  sp_flows            3
  bk_flows            2
  un_flows            0
  max_len         14280
  quantum           976

ouch! FTTW... traditional qos might be your only hope luke. them's up's be not on the up-and-up...

interested to see another (dslreport) test, during a quiet time ( late-late night or early morning )...

Thanks, any way to cap my upload speed while keeping the sqm? And any suggestion for the Link layer adaptation setting for my type of cennection?

forget link layer adaption... ;

  • rule out RFI
  • ditch SQM... for a simpler qos

( but given the test above... you could 'squeeze' things a little more perhaps... for testing sake )

sqm.@queue[0].download -> 27000-29000
sqm.@queue[0].upload -> 830-1200

I see you like dslreport results so here one with luci qos http://www.dslreports.com/speedtest/64793223 it seems worse to me, bufferbloat in downloads too.
Probably stick for sqm even if it's more placebo then useful.

1 Like

interesting... well that rules out RFI ... ( as expected )

The test failed many times at 27000/830

Well, the way SQM works it requires stable bandwidth on up- and downlink, it can unfortunately not deal well with variable bandwidth... Some users get away with setting a bandwidth lower than the contracted rate that their link reliably delivers, but judging from your tests that seems not really feasible for you. But I do note thart even your old dsl speedtest has a bad upload....