Skip to contentSkip to conversation

Bandwidth throttling/limiting for all traffic

LARTC Darryl Cording 3 participants 7 messages Nov 18, 2004

lore
Share & tools
lore
#0 lore raw

Hi,

I have a rather simple problem I have to solve, well I thought it would be simple, I've run into a problem. I think I must be missing something fundamental.

I am trying to build a Linux router that simply throttles everything down to certain bandwidths. That is, no priority queuing ...etc, just slow all traffic down to the specified rates, which are 64,128,256,512 kbit. We want to setup the bandwidth on the router, then pump a known load from an application through the router and see how that behaves over various link speeds. (we are in a lab environment)

Using kernel 2.6.5-1.358 I simply used these commands from Stef's version of tc from docum.org, module sch_htb is loaded.

tc qdisc add dev eth0 root handle 10: htb
tc class add dev eth0 parent 10: classid 10:1 htb rate 64kbit ceil 64kit

I thought that would do the job, but my testing from uploading files via ftp to another box, show nothing gets shaped.

I think I've missed something here and need a clue. Alternatively can anyone suggest any programs that are designed for doing this type of work.

Thanks
Darryl

_______________________________________________
LARTC mailing list / [email protected]
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/

#1 lore raw
Quoted 2 lines · Darryl Cording

Hi,

<snip>

Quoted 6 lines

I am trying to build a Linux router that simply throttles everything down to certain bandwidths. That is, no priority queuing ...etc, just slow all traffic down to the specified rates, which are 64,128,256,512 kbit. We want to setup the bandwidth on the router, then pump a known load from an application through the router and see how that behaves over various link speeds. (we are in a lab environment)

Using the htb qdisc with ceil equal to rate ought to accomplish this.

Quoted 5 lines

Using kernel 2.6.5-1.358 I simply used these commands from Stef's version of tc from docum.org, module sch_htb is loaded.

tc qdisc add dev eth0 root handle 10: htb
tc class add dev eth0 parent 10: classid 10:1 htb rate 64kbit ceil 64kit

You need to classify your traffic.

tc qdisc add dev eth0 root handle 10: htb default 10
tc class add dev eth0 parent 10: classid 10:1 htb rate 64kbit ceil 64kit

If you add the default parameter, htb will put any unclassified into classid 10:1.

Quoted 2 lines

I thought that would do the job, but my testing from uploading files via ftp to another box, show nothing gets shaped.

Right, because it wasn't classified.

Signature · 8 lines
Jason Boxman
Perl Programmer / *NIX Systems Administrator
Shimberg Center for Affordable Housing | University of Florida
http://edseek.com/ - Linux and FOSS stuff

_______________________________________________
LARTC mailing list / [email protected]
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
#2 lore raw
In reply to Darryl Cording

Thanks for the feedback Jason,

Quoted 12 lines · Jason Boxman

On Wednesday 17 November 2004 22:29, Darryl Cording wrote:

<snip>

Quoted 6 lines

I am trying to build a Linux router that simply throttles everything down to certain bandwidths. That is, no priority queuing ...etc, just slow all traffic down to the specified rates, which are 64,128,256,512 kbit. We want to setup the bandwidth on the router, then pump a known load from an application through the router and see how that behaves over various link speeds. (we are in a lab environment)

Quoted 1 line

Using the htb qdisc with ceil equal to rate ought to accomplish this.

I thought the same thing, so far so good.

Quoted 8 lines
Quoted 6 lines

Using kernel 2.6.5-1.358 I simply used these commands from Stef's version of tc from docum.org, module sch_htb is loaded.

tc qdisc add dev eth0 root handle 10: htb
tc class add dev eth0 parent 10: classid 10:1 htb rate 64kbit ceil 64kit

Quoted 7 lines

You need to classify your traffic.

tc qdisc add dev eth0 root handle 10: htb default 10
tc class add dev eth0 parent 10: classid 10:1 htb rate 64kbit ceil 64kit

If you add the default parameter, htb will put any unclassified into classid 10:1.

That makes sense.

Quoted 8 lines
Quoted 4 lines

I thought that would do the job, but my testing from uploading files via ftp to another box, show nothing gets shaped.

Right, because it wasn't classified.

Ok, so I have to classify my traffic before this will route them throu the qdisc. Are you taking about classifying via iptables?? I thought that was optional, more for filtering ...etc.

regards
darryl

_______________________________________________
LARTC mailing list / [email protected]
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/

#3 lore raw
In reply to Darryl Cording
Quoted 10 lines · Darryl Cording
Quoted 3 lines

Right, because it wasn't classified.

Ok, so I have to classify my traffic before this will route them throu the qdisc. Are you taking about classifying via iptables?? I thought that was optional, more for filtering ...etc.

regards
darryl

I was getting confused with the terminology. I was thinking filtering was meaning something else when "tc filter ..." actually does the classifying. I was also assuming that if you don't specify any packet attributes to filter on it would just catch everything, seems it's the opposite.

OK, so I used this;

tc qdisc add dev eth0 root handle 10: htb default 10
tc class add dev eth0 parent 10: classid 10:1 htb rate 64kbit ceil 64kbit
and have been experimenting with classifying like this,
tc filter add dev eth0 parent 10: protocol ip prio 1 u32 match ip protocol 0 0xff
tc filter add dev eth0 parent 10: protocol ip prio 1 u32 match ip protocol 6 0xff

But it seems my ftp transfers are not being shaped, in fact, lol, they are going faster from when I first started experimenting. So it's not matching correctly. I just want to shape everything going past the NIC's. I thought that if I could classify the entire ip protocol or the tcp protocol that would shape the bulk of the traffic ? I was hoping not having to get down to specifying ports, but find a simple way to shape all outgoing traffic on a multi-homed host.

Am I on the right track?

And what worries me also now, is that the application traffic coming over the wire will be DCOM, which I think is rpc based, another world of hurt if I have to figure out what ports that app is using.

thanks for any tips,
darryl

_______________________________________________
LARTC mailing list / [email protected]
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/

#4 lore raw
In reply to Darryl Cording
Quoted 11 lines · Darryl Cording

But it seems my ftp transfers are not being shaped, in fact, lol, they are going faster from when I first started experimenting. So it's not matching correctly. I just want to shape everything going past the NIC's. I thought that if I could classify the entire ip protocol or the tcp protocol that would shape the bulk of the traffic ? I was hoping not having to get down to specifying ports, but find a simple way to shape all outgoing traffic on a multi-homed host.

Am I on the right track?

I guess I could match by ip addresses but I hoping for a simpler way to match everthing.

_______________________________________________
LARTC mailing list / [email protected]
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/

#5 lore raw
In reply to Darryl Cording
Quoted 11 lines · Darryl Cording
Quoted 9 lines · Darryl Cording

But it seems my ftp transfers are not being shaped, in fact, lol, they are going faster from when I first started experimenting. So it's not matching correctly. I just want to shape everything going past the NIC's. I thought that if I could classify the entire ip protocol or the tcp protocol that would shape the bulk of the traffic ? I was hoping not having to get down to specifying ports, but find a simple way to shape all outgoing traffic on a multi-homed host.

Quoted 4 lines
Quoted 2 lines

Am I on the right track?

I guess I could match by ip addresses but I hoping for a simpler way to match everthing.

You were close - default refers to the number after the :

10: is short for 10:0 .

tc qdisc add dev eth0 root handle 10: htb default 1
tc class add dev eth0 parent 10: classid 10:1 htb rate 64kbit ceil 64kbit

should shape everything.

Andy.

_______________________________________________
LARTC mailing list / [email protected]
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/

#6 lore raw
In reply to Darryl Cording
Quoted 8 lines · Andy Furniss
Quoted 3 lines · Darryl Cording
Quoted 1 line

I guess I could match by ip addresses but I hoping for a simpler way to match everthing.

You were close - default refers to the number after the :

I guess I am still struggling with the syntax of tc.

Quoted 2 lines

10: is short for 10:0 .

Thanks for that clue.

Quoted 5 lines

tc qdisc add dev eth0 root handle 10: htb default 1
tc class add dev eth0 parent 10: classid 10:1 htb rate 64kbit ceil 64kbit

should shape everything.

And in fact it does. :-D
Thanks for the help.

Quoted 8 lines

Andy.

_______________________________________________
LARTC mailing list / [email protected]
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/

Signature · 14 lines
Ascend Business Solutions Pty Ltd
Level 3, 232 Adelaide St, Brisbane, Qld, 4000
Phone (07) 3223 5555 Fax (07) 3223 5500
www.ascend.net.au

This email is from the offices of Ascend Business Solutions Pty Ltd,
Software Testing Specialists, Brisbane, Queensland, Australia. This email
contains privileged and confidential information. It is intended for the
recipient only. Should this email be received in error please notify this
office immediately by reply email and delete this email from your inbox.

_______________________________________________
LARTC mailing list / [email protected]
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/