All of lore.kernel.org
 help / color / mirror / Atom feed
* [LARTC] CBQ_bandwidth
@ 2003-08-20 10:50 Raghuveer
  2003-08-20 12:22 ` Stef Coene
                   ` (8 more replies)
  0 siblings, 9 replies; 10+ messages in thread
From: Raghuveer @ 2003-08-20 10:50 UTC (permalink / raw)
  To: lartc


Hi Stef,

Reffering to page < http://www.docum.org/stef.coene/qos/faq/cache/7.html 
 > I found a para as follows,

" CBQ is not always as accurate as it should be. See docum.org on the 
test page for some tests with bounded classes. The algorithm used by cbq 
is very "link sensitive". It calculates the link idle time and for that 
it needs the real link bandwidth, average packet size and some other 
parameters. But there calculations are not always accurate. So it's 
possible that you will get inaccurate results if you use CBQ to limit 
the traffic in a class.  "

Here Iam slightly confused with real link bandwidth, Can I understand 
that while creating CBQ we need interface bandwidth (ethtool, mii-diag) 
and and while assigning the classes real link bandwidth is required. An 
example will definately help me. Can you pls suggest a way to find real 
link bandwidth....? As ethtool and mii-diag will only get the interface 
bandwidth.

Regards
-Raghu

_______________________________________________
LARTC mailing list / LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [LARTC] CBQ_bandwidth
  2003-08-20 10:50 [LARTC] CBQ_bandwidth Raghuveer
@ 2003-08-20 12:22 ` Stef Coene
  2003-08-20 12:57 ` Raghuveer
                   ` (7 subsequent siblings)
  8 siblings, 0 replies; 10+ messages in thread
From: Stef Coene @ 2003-08-20 12:22 UTC (permalink / raw)
  To: lartc

On Wednesday 20 August 2003 12:50, Raghuveer wrote:
> Hi Stef,
>
> Reffering to page < http://www.docum.org/stef.coene/qos/faq/cache/7.html
>
>  > I found a para as follows,
>
> " CBQ is not always as accurate as it should be. See docum.org on the
> test page for some tests with bounded classes. The algorithm used by cbq
> is very "link sensitive". It calculates the link idle time and for that
> it needs the real link bandwidth, average packet size and some other
> parameters. But there calculations are not always accurate. So it's
> possible that you will get inaccurate results if you use CBQ to limit
> the traffic in a class.  "
>
> Here Iam slightly confused with real link bandwidth, Can I understand
> that while creating CBQ we need interface bandwidth (ethtool, mii-diag)
> and and while assigning the classes real link bandwidth is required. An
> example will definately help me. Can you pls suggest a way to find real
> link bandwidth....? As ethtool and mii-diag will only get the interface
> bandwidth.
The bandwidth parameter is the real link bandwidth like ethtook or mii-diag 
will show you.
If you want to limit all traffic to the link bandwidth, you have to create a 
class that's bounded and with rate = link bandwidth.  You can create 
non-bounded sub classes but the bandwidth will be limit by the bounded class 
to the link bandwidth.

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/

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [LARTC] CBQ_bandwidth
  2003-08-20 10:50 [LARTC] CBQ_bandwidth Raghuveer
  2003-08-20 12:22 ` Stef Coene
@ 2003-08-20 12:57 ` Raghuveer
  2003-08-20 15:09 ` Stef Coene
                   ` (6 subsequent siblings)
  8 siblings, 0 replies; 10+ messages in thread
From: Raghuveer @ 2003-08-20 12:57 UTC (permalink / raw)
  To: lartc

Stef Coene wrote:

>On Wednesday 20 August 2003 12:50, Raghuveer wrote:
>  
>
>>Hi Stef,
>>
>>Reffering to page < http://www.docum.org/stef.coene/qos/faq/cache/7.html
>>
>> > I found a para as follows,
>>
>>" CBQ is not always as accurate as it should be. See docum.org on the
>>test page for some tests with bounded classes. The algorithm used by cbq
>>is very "link sensitive". It calculates the link idle time and for that
>>it needs the real link bandwidth, average packet size and some other
>>parameters. But there calculations are not always accurate. So it's
>>possible that you will get inaccurate results if you use CBQ to limit
>>the traffic in a class.  "
>>
>>Here Iam slightly confused with real link bandwidth, Can I understand
>>that while creating CBQ we need interface bandwidth (ethtool, mii-diag)
>>and and while assigning the classes real link bandwidth is required. An
>>example will definately help me. Can you pls suggest a way to find real
>>link bandwidth....? As ethtool and mii-diag will only get the interface
>>bandwidth.
>>    
>>
>The bandwidth parameter is the real link bandwidth like ethtook or mii-diag 
>will show you.
>If you want to limit all traffic to the link bandwidth, you have to create a 
>class that's bounded and with rate = link bandwidth.  You can create 
>non-bounded sub classes but the bandwidth will be limit by the bounded class 
>to the link bandwidth.
>  
>
What is confusing me is, there is a bandwidth provided by ISP (512Kbits) 
and one ethernet capacity(100Mbits), so which one can we call as real 
link bandwidth.  What is NIC bandwidth....is it ethernet bandwidth or 
ISP bandwidth....?
    Lan------->eth1-----------eth0---------->Internet
Now at eth0 I have ethernet device bandwidth as 100Mbits and my ISP 
provides 512Kbits bandwidth. so if I want to do egress traffic control 
at both eth0 and eth1, what bandwidth I should consider...? My eth1 
ethernet device bandwidth is 100Mbits.

Regards
-Raghu

>Stef
>
>  
>


_______________________________________________
LARTC mailing list / LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [LARTC] CBQ_bandwidth
  2003-08-20 10:50 [LARTC] CBQ_bandwidth Raghuveer
  2003-08-20 12:22 ` Stef Coene
  2003-08-20 12:57 ` Raghuveer
@ 2003-08-20 15:09 ` Stef Coene
  2003-08-21  4:57 ` Raghuveer
                   ` (5 subsequent siblings)
  8 siblings, 0 replies; 10+ messages in thread
From: Stef Coene @ 2003-08-20 15:09 UTC (permalink / raw)
  To: lartc

On Wednesday 20 August 2003 14:58, Raghuveer wrote:
> What is confusing me is, there is a bandwidth provided by ISP (512Kbits)
> and one ethernet capacity(100Mbits), so which one can we call as real
> link bandwidth.  What is NIC bandwidth....is it ethernet bandwidth or
> ISP bandwidth....?
>     Lan------->eth1-----------eth0---------->Internet
> Now at eth0 I have ethernet device bandwidth as 100Mbits and my ISP
> provides 512Kbits bandwidth. so if I want to do egress traffic control
> at both eth0 and eth1, what bandwidth I should consider...? My eth1
> ethernet device bandwidth is 100Mbits.
What bout this :

for all cbq commands : bandwidth 100mbit
eth0
  cbq qdisc
    cbq class rate = 512kbit, bounded
       cbq class 1, rate < 512kbit
       cbq class ..., rate < 512kbit
       cbq class x, rate < 512kbit
So all traffic from class 1 ... x togehter is bounded to 512kbit.  

eth1
  cbq qdisc
    cbq class rate = 100mbit, bounded
       cbq class 1, rate 512kbit bounded
          cbq class 10, rate < 512kbit
          cbq class ..., rate < 512kbit
          cbq class x, rate < 512kbit
       cbq class 2, rate 99,5Mbit
          cbq class 20, rate < 99,5Mbit
          cbq class ..., rate < 99,5Mbit
          cbq class x, rate < 99,5Mbit

Class 1 is for all traffic from internet -> LAN
Class 2 is for all traffic from shaper -> LAN

And if you really want to be sure it's working, you should take 500kbit.  So 
YOU are the bottleneck and in control of the link and not the modem.

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/

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [LARTC] CBQ_bandwidth
  2003-08-20 10:50 [LARTC] CBQ_bandwidth Raghuveer
                   ` (2 preceding siblings ...)
  2003-08-20 15:09 ` Stef Coene
@ 2003-08-21  4:57 ` Raghuveer
  2003-08-21 18:20 ` Stef Coene
                   ` (4 subsequent siblings)
  8 siblings, 0 replies; 10+ messages in thread
From: Raghuveer @ 2003-08-21  4:57 UTC (permalink / raw)
  To: lartc

Stef Coene wrote:

>On Wednesday 20 August 2003 14:58, Raghuveer wrote:
>  
>
>>What is confusing me is, there is a bandwidth provided by ISP (512Kbits)
>>and one ethernet capacity(100Mbits), so which one can we call as real
>>link bandwidth.  What is NIC bandwidth....is it ethernet bandwidth or
>>ISP bandwidth....?
>>    Lan------->eth1-----------eth0---------->Internet
>>Now at eth0 I have ethernet device bandwidth as 100Mbits and my ISP
>>provides 512Kbits bandwidth. so if I want to do egress traffic control
>>at both eth0 and eth1, what bandwidth I should consider...? My eth1
>>ethernet device bandwidth is 100Mbits.
>>    
>>
>What bout this :
>
>for all cbq commands : bandwidth 100mbit
>eth0
>  cbq qdisc
>    cbq class rate = 512kbit, bounded
>       cbq class 1, rate < 512kbit
>       cbq class ..., rate < 512kbit
>       cbq class x, rate < 512kbit
>So all traffic from class 1 ... x togehter is bounded to 512kbit.  
>
>eth1
>  cbq qdisc
>    cbq class rate = 100mbit, bounded
>       cbq class 1, rate 512kbit bounded
>          cbq class 10, rate < 512kbit
>          cbq class ..., rate < 512kbit
>          cbq class x, rate < 512kbit
>       cbq class 2, rate 99,5Mbit
>          cbq class 20, rate < 99,5Mbit
>          cbq class ..., rate < 99,5Mbit
>          cbq class x, rate < 99,5Mbit
>
>Class 1 is for all traffic from internet -> LAN
>Class 2 is for all traffic from shaper -> LAN
>
>And if you really want to be sure it's working, you should take 500kbit.  So 
>YOU are the bottleneck and in control of the link and not the modem.
>  
>
Thanks alot Stef, Its very clear to me now.
If I have ADSL with different incoming and outgoing rates. For egress 
shaping at WAN(eth0) interface, Whether should I use outgoing rates or 
combined rates (incomming + outgoing). Will it make any big difference 
if I use combined rates in an asymmetric link like cabel modem, ADSL etc.

Regards
-Raghu

>Stef
>
>  
>


_______________________________________________
LARTC mailing list / LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [LARTC] CBQ_bandwidth
  2003-08-20 10:50 [LARTC] CBQ_bandwidth Raghuveer
                   ` (3 preceding siblings ...)
  2003-08-21  4:57 ` Raghuveer
@ 2003-08-21 18:20 ` Stef Coene
  2003-08-22  5:29 ` Raghuveer
                   ` (3 subsequent siblings)
  8 siblings, 0 replies; 10+ messages in thread
From: Stef Coene @ 2003-08-21 18:20 UTC (permalink / raw)
  To: lartc

On Thursday 21 August 2003 06:58, Raghuveer wrote:
> Thanks alot Stef, Its very clear to me now.
> If I have ADSL with different incoming and outgoing rates. For egress
> shaping at WAN(eth0) interface, Whether should I use outgoing rates or
> combined rates (incomming + outgoing). Will it make any big difference
> if I use combined rates in an asymmetric link like cabel modem, ADSL etc.
At eth0, you should take outgoing rates.  And this can be assymetric from 
incoming rates.  Why not ?  
I didn't knew your incoming and outgoing rates so I just took 512kbit for 
incoming and outgoing.

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/

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [LARTC] CBQ_bandwidth
  2003-08-20 10:50 [LARTC] CBQ_bandwidth Raghuveer
                   ` (4 preceding siblings ...)
  2003-08-21 18:20 ` Stef Coene
@ 2003-08-22  5:29 ` Raghuveer
  2003-08-22  8:36 ` Stef Coene
                   ` (2 subsequent siblings)
  8 siblings, 0 replies; 10+ messages in thread
From: Raghuveer @ 2003-08-22  5:29 UTC (permalink / raw)
  To: lartc

Stef Coene wrote:

>On Thursday 21 August 2003 06:58, Raghuveer wrote:
>  
>
>>Thanks alot Stef, Its very clear to me now.
>>If I have ADSL with different incoming and outgoing rates. For egress
>>shaping at WAN(eth0) interface, Whether should I use outgoing rates or
>>combined rates (incomming + outgoing). Will it make any big difference
>>if I use combined rates in an asymmetric link like cabel modem, ADSL etc.
>>    
>>
>At eth0, you should take outgoing rates.  And this can be assymetric from 
>incoming rates.  Why not ?  
>I didn't knew your incoming and outgoing rates so I just took 512kbit for 
>incoming and outgoing.
>  
>
How can I get the incomming and outgoing rates seperately ...?  

Regards
-Raghu

>Stef
>
>  
>


_______________________________________________
LARTC mailing list / LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [LARTC] CBQ_bandwidth
  2003-08-20 10:50 [LARTC] CBQ_bandwidth Raghuveer
                   ` (5 preceding siblings ...)
  2003-08-22  5:29 ` Raghuveer
@ 2003-08-22  8:36 ` Stef Coene
  2003-08-22  9:22 ` Raghuveer
  2003-08-22 11:39 ` Stef Coene
  8 siblings, 0 replies; 10+ messages in thread
From: Stef Coene @ 2003-08-22  8:36 UTC (permalink / raw)
  To: lartc

On Friday 22 August 2003 07:17, Raghuveer wrote:
> Stef Coene wrote:
> >On Thursday 21 August 2003 06:58, Raghuveer wrote:
> >>Thanks alot Stef, Its very clear to me now.
> >>If I have ADSL with different incoming and outgoing rates. For egress
> >>shaping at WAN(eth0) interface, Whether should I use outgoing rates or
> >>combined rates (incomming + outgoing). Will it make any big difference
> >>if I use combined rates in an asymmetric link like cabel modem, ADSL etc.
> >
> >At eth0, you should take outgoing rates.  And this can be assymetric from
> >incoming rates.  Why not ?
> >I didn't knew your incoming and outgoing rates so I just took 512kbit for
> >incoming and outgoing.
>
> How can I get the incomming and outgoing rates seperately ...?
What do you mean?  How to configure htb or how to get these numbers from you 
ISP?

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/

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [LARTC] CBQ_bandwidth
  2003-08-20 10:50 [LARTC] CBQ_bandwidth Raghuveer
                   ` (6 preceding siblings ...)
  2003-08-22  8:36 ` Stef Coene
@ 2003-08-22  9:22 ` Raghuveer
  2003-08-22 11:39 ` Stef Coene
  8 siblings, 0 replies; 10+ messages in thread
From: Raghuveer @ 2003-08-22  9:22 UTC (permalink / raw)
  To: lartc

Stef Coene wrote:

>On Friday 22 August 2003 07:17, Raghuveer wrote:
>  
>
>>Stef Coene wrote:
>>    
>>
>>>On Thursday 21 August 2003 06:58, Raghuveer wrote:
>>>      
>>>
>>>>Thanks alot Stef, Its very clear to me now.
>>>>If I have ADSL with different incoming and outgoing rates. For egress
>>>>shaping at WAN(eth0) interface, Whether should I use outgoing rates or
>>>>combined rates (incomming + outgoing). Will it make any big difference
>>>>if I use combined rates in an asymmetric link like cabel modem, ADSL etc.
>>>>        
>>>>
>>>At eth0, you should take outgoing rates.  And this can be assymetric from
>>>incoming rates.  Why not ?
>>>I didn't knew your incoming and outgoing rates so I just took 512kbit for
>>>incoming and outgoing.
>>>      
>>>
>>How can I get the incomming and outgoing rates seperately ...?
>>    
>>
>What do you mean?  How to configure htb or how to get these numbers from you 
>ISP?
>
Actually I wanted to know whether is there any way to get the incoming 
and outgoing rates using some tool.  

Regards
-Raghu

>
>Stef
>
>  
>


_______________________________________________
LARTC mailing list / LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [LARTC] CBQ_bandwidth
  2003-08-20 10:50 [LARTC] CBQ_bandwidth Raghuveer
                   ` (7 preceding siblings ...)
  2003-08-22  9:22 ` Raghuveer
@ 2003-08-22 11:39 ` Stef Coene
  8 siblings, 0 replies; 10+ messages in thread
From: Stef Coene @ 2003-08-22 11:39 UTC (permalink / raw)
  To: lartc

On Friday 22 August 2003 11:10, Raghuveer wrote:
> Actually I wanted to know whether is there any way to get the incoming
> and outgoing rates using some tool.
The rate is (most of the time) limited in the modem.  And there is no way you 
can find this.  You can try to upload / download a file to / from a fast ftp 
server and monitor the bandwidth you get, but that's not very accurate.
Even if you don't know the bandwidth, you can shape the traffic. 

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/

^ permalink raw reply	[flat|nested] 10+ messages in thread

end of thread, other threads:[~2003-08-22 11:39 UTC | newest]

Thread overview: 10+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2003-08-20 10:50 [LARTC] CBQ_bandwidth Raghuveer
2003-08-20 12:22 ` Stef Coene
2003-08-20 12:57 ` Raghuveer
2003-08-20 15:09 ` Stef Coene
2003-08-21  4:57 ` Raghuveer
2003-08-21 18:20 ` Stef Coene
2003-08-22  5:29 ` Raghuveer
2003-08-22  8:36 ` Stef Coene
2003-08-22  9:22 ` Raghuveer
2003-08-22 11:39 ` Stef Coene

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.