Hi,
I tried the recommendations of people on this list and tried tweaking
the QUANTUM values in my HTB configuration, but still have the problem
of the excess bandwidth not being shared proportionately between
different pools.
Am enclosing the set of tc commands I''m using to generate the pools.
The values have been desperately tweaked for some 4 hours, so some of
them may look a bit funny.
Basically what I''m trying to achieve is that when PCs in (e.g.) 1:4
and 1:5 try downloading at top speed, they should get bandwidths in
ratio of their rates, in a 64:100 proportion. However, what actually
happens is that the 256Kbps downlink is shared equally between both
the classes.
Have been beating my head against this wall for days now, any pointers
much appreciated.
Regards,
-- Raju
--
Raj Mathur raju@kandalaya.org http://kandalaya.org/
GPG: 78D4 FC67 367F 40E2 0DD5 0FEF C968 D0EF CC68 D17F
It is the mind that moves
#!/bin/sh
#
# Test manual HTB configuration from lartc
#
# Root qdisc
#
tc qdisc del dev eth1 root
tc qdisc add dev eth1 root handle 1: htb default 4
#
# Interface rate
#
tc class add dev eth1 parent 1: classid 1:3 htb rate 256Kbit ceil 256Kbit prio 5
burst 60k cburst 60k
#
# Rate limiting classes
#
tc class add dev eth1 parent 1:3 classid 1:4 htb rate 64Kbit ceil 256Kbit prio 5
quantum 30000 burst 30000 cburst 30000
tc class add dev eth1 parent 1:3 classid 1:5 htb rate 100Kbit ceil 256Kbit prio
5 quantum 60000 burst 60000 cburst 60000
tc class add dev eth1 parent 1:3 classid 1:6 htb rate 32Kbit ceil 128Kbit prio 5
quantum 1500 burst 1500 cburst 1500
#
# Sharing qdiscs
#
tc qdisc add dev eth1 parent 1:4 handle 4: sfq perturb 10 quantum 30000
tc qdisc add dev eth1 parent 1:5 handle 5: sfq perturb 10 quantum 4500
tc qdisc add dev eth1 parent 1:6 handle 6: sfq perturb 10 quantum 1500
#
# Mark match
#
tc filter add dev eth1 protocol ip prio 5 handle 65537 fw flowid 1:4
tc filter add dev eth1 protocol ip prio 5 handle 65538 fw flowid 1:5
tc filter add dev eth1 protocol ip prio 5 handle 65539 fw flowid 1:6
_______________________________________________
LARTC mailing list / LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
Martin A. Brown
2003-Aug-01 15:13 UTC
Re: Proportionate sharing of excess bandwidth: still no go
Hi Raj, : I tried the recommendations of people on this list and tried tweaking : the QUANTUM values in my HTB configuration, but still have the problem : of the excess bandwidth not being shared proportionately between : different pools. Could you tell us which interface is your "transmit to the Internet" interface? I speculate that you are shaping your upload bandwidth, not your download bandwidth. If my speculation is correct, perhaps you could try using the same traffic control configuration on both devices (eth0 ? and eth1). What happens then? : Am enclosing the set of tc commands I''m using to generate the pools. : The values have been desperately tweaked for some 4 hours, so some of : them may look a bit funny. Your quantum setting seems awfully high to me. I''d recommend removing any reference to quantum before trying again. : Basically what I''m trying to achieve is that when PCs in (e.g.) 1:4 and : 1:5 try downloading at top speed, they should get bandwidths in ratio : of their rates, in a 64:100 proportion. However, what actually happens : is that the 256Kbps downlink is shared equally between both the : classes. Seems reasonable. Best of luck, -Martin -- Martin A. Brown --- SecurePipe, Inc. --- mabrown@securepipe.com _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
Raj Mathur
2003-Aug-01 15:22 UTC
Re: Proportionate sharing of excess bandwidth: still no go
Hi Martin,>>>>> "Martin" == Martin A Brown <mabrown-lartc@securepipe.com> writes:Raj> : I tried the recommendations of people on this list Raj> and tried tweaking : the QUANTUM values in my HTB Raj> configuration, but still have the problem : of the excess Raj> bandwidth not being shared proportionately between : Raj> different pools. Martin> Could you tell us which interface is your "transmit to the Martin> Internet" interface? Martin> I speculate that you are shaping your upload bandwidth, Martin> not your download bandwidth. If my speculation is Martin> correct, perhaps you could try using the same traffic Martin> control configuration on both devices (eth0 ? and eth1). Martin> What happens then? This is for the `internal'' (LAN) interface, eth1, which shapes incoming traffic. I''m running similar commands on the `external'' (WAN) interface, eth0, for shaping outbound traffic. Not too concerned about aspect, so omitted in the interests of brevity. Raj> : Am enclosing the set of tc commands I''m using to generate Raj> the pools. : The values have been desperately tweaked for Raj> some 4 hours, so some of : them may look a bit funny. Martin> Your quantum setting seems awfully high to me. I''d Martin> recommend removing any reference to quantum before trying Martin> again. Originally it was something like 1500, 3000 and 6000 :) Like I said, I was fooling around with the values to see if it made any difference at all -- as you can guess, it didn''t. Martin> [snip] Regards, -- Raju -- Raj Mathur raju@kandalaya.org http://kandalaya.org/ GPG: 78D4 FC67 367F 40E2 0DD5 0FEF C968 D0EF CC68 D17F It is the mind that moves _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
Stef Coene
2003-Aug-01 15:41 UTC
Re: Proportionate sharing of excess bandwidth: still no go
On Friday 01 August 2003 17:13, Martin A. Brown wrote:> Hi Raj, > > : I tried the recommendations of people on this list and tried tweaking > : the QUANTUM values in my HTB configuration, but still have the problem > : of the excess bandwidth not being shared proportionately between > : different pools. > > Could you tell us which interface is your "transmit to the Internet" > interface? > > I speculate that you are shaping your upload bandwidth, not your download > bandwidth. If my speculation is correct, perhaps you could try using the > same traffic control configuration on both devices (eth0 ? and eth1). > What happens then? > > : Am enclosing the set of tc commands I''m using to generate the pools. > : The values have been desperately tweaked for some 4 hours, so some of > : them may look a bit funny. > > Your quantum setting seems awfully high to me. I''d recommend removing any > reference to quantum before trying again.Better, remove burst, cburst and quantum. And remove the sfq qdiscs and add 3 fifo qdiscs. Stef -- stef.coene@docum.org "Using Linux as bandwidth manager" http://www.docum.org/ #lartc @ irc.oftc.net _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
Raj Mathur
2003-Aug-01 15:56 UTC
Re: Proportionate sharing of excess bandwidth: still no go
Hi Stef,>>>>> "Stef" == Stef Coene <stef.coene@docum.org> writes:Stef> [snip] Stef> Better, remove burst, cburst and quantum. And remove the Stef> sfq qdiscs and add 3 fifo qdiscs. Can do that, but will it permit the PCs in each pool to share the bandwidth available to that pool equitably? Regards, -- Raju -- Raj Mathur raju@kandalaya.org http://kandalaya.org/ GPG: 78D4 FC67 367F 40E2 0DD5 0FEF C968 D0EF CC68 D17F It is the mind that moves _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
Stef Coene
2003-Aug-02 08:08 UTC
Re: Proportionate sharing of excess bandwidth: still no go
On Friday 01 August 2003 17:56, Raj Mathur wrote:> Hi Stef, > > >>>>> "Stef" == Stef Coene <stef.coene@docum.org> writes: > > Stef> [snip] > > Stef> Better, remove burst, cburst and quantum. And remove the > Stef> sfq qdiscs and add 3 fifo qdiscs. > > Can do that, but will it permit the PCs in each pool to share the > bandwidth available to that pool equitably?No. But the sfq will do that neither. You need the esfq qdisc to do that. See http://www.docum.org/stef.coene/qos/faq/cache/38.html Stef -- stef.coene@docum.org "Using Linux as bandwidth manager" http://www.docum.org/ #lartc @ irc.oftc.net _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/