I have been using Gargoyle for years and decided to try something new. What kept me using Gargoyle is the Active Congestion Control QoS feature that automatically keeps ping times at a user defined setting on my wireless connection, where the download bandwidth changes wildly with the weather. But I read that SQM has a "autorate-ingress" feature that aims to achieve the same thing.
Today I sysupgraded to OpenWrt on my Linksys WRT1900ACS and have been learning about and testing the SQM feature according to the man page.
My current settings are
root@OpenWrt:~# cat /etc/config/sqm
config queue
option debug_logging '0'
option verbosity '5'
option squash_ingress '1'
option ingress_ecn 'ECN'
option egress_ecn 'NOECN'
option interface 'eth1.2'
option upload '4000'
option qdisc 'cake'
option qdisc_advanced '1'
option qdisc_really_really_advanced '1'
option enabled '1'
option linklayer 'ethernet'
option download '70000'
option overhead '44'
option script 'piece_of_cake.qos'
option squash_dscp '1'
option iqdisc_opts 'autorate-ingress rtt 40ms'
When I have Windows pinging 1.1.1.1, I get good ping about 30-40ms. Then I constantly run ookla speedtests and SQM keeps ping the same range. But when I start a download in steam, pings go upto 100-200ms. What can I do to stop steam increasing ping times as much? The ookla speedtest download result is always ~50% lower than steam so it is like steam is bypassing SQM? When I turn off autorate-ingress steam doesn't cause pings to increase but don't get the benefits of dynamic bandwidth estimation on wireless links.
Also I don't really understand what the option "rtt 40ms" does. I can see in the output below that "interval" is now 40ms but what does interval mean? When I set rtt to 200ms, nothing seems to change regarding Windows 1.1.1.1 ping times.
root@OpenWrt:~# tc -s qdisc show dev ifb4eth1.2
qdisc cake 812b: root refcnt 2 bandwidth 29432Kbit autorate-ingress besteffort triple-isolate nonat wash no-ack-filter split-gso rtt 40.0ms noatm overhead 44
Sent 361346 bytes 3795 pkt (dropped 0, overlimits 3553 requeues 0)
backlog 0b 0p requeues 0
memory used: 7296b of 4Mb
capacity estimate: 31395Kbit
min/max network layer size: 46 / 1352
min/max overhead-adjusted size: 90 / 1396
average network hdr offset: 14
Tin 0
thresh 29432Kbit
target 2.0ms
interval 40.0ms
pk_delay 314us
av_delay 45us
sp_delay 3us
backlog 0b
pkts 3795
bytes 361346
way_inds 0
way_miss 17
way_cols 0
drops 0
marks 0
ack_drop 0
sp_flows 6
bk_flows 1
un_flows 0
max_len 1366
quantum 898