* Re: Fwd: [LARTC] HTB shaping & borrowing info
2006-05-27 22:11 Fwd: [LARTC] HTB shaping & borrowing info Stefano Mainardi
@ 2006-05-27 23:24 ` Andreas Klauer
2006-05-28 0:04 ` Stefano Mainardi
` (10 subsequent siblings)
11 siblings, 0 replies; 13+ messages in thread
From: Andreas Klauer @ 2006-05-27 23:24 UTC (permalink / raw)
To: lartc
On Sun, May 28, 2006 at 12:11:10AM +0200, Stefano Mainardi wrote:
> >> If B don't make traffic, 7/8 of 20Mb/s must be assigned to A and all the
> >> rest at B
>
> Sorry, "all the rest at A" :)
So, in other words, A is allowed to take bandwidth from B.
B and C stick to their bandwidth limits.
A tree like this could probably accomplish this:
HTB qdisc
|
\--- HTB root class (100mbit)
|
\--- HTB class (90mbit|90mbit)
| |
| \--- HTB class A (70mbit|90mbit)
| \--- HTB class B (20mbit|20mbit)
|
\--- HTB class C (10mbit|10mbit)
This way, C and B never borrow any bandwidth (as they have rate=ceil),
and if A borrows, it will be from B, as the parent class (which has
rate=ceil as well) will never borrow from C.
> We have tested this script with CEIL=RATE, and CEIL\x100Mbit, but i view that
> the data-rate calculated for each PC is not proportional to the traffic
> assigned at Firewall.
HTB expects to be able to use the full specified rate at any point
of time, so you probably should use something lower than 100mbit
as a base value. Even in 100mbit networks, you never actually get
this rate, due to overhead, collisions, etc.
Other than that, are there really just these three classes of traffic
going out on eth1? The setup should work, as long as the classification
is working properly.
Regards
Andreas Klauer
_______________________________________________
LARTC mailing list
LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
^ permalink raw reply [flat|nested] 13+ messages in thread* Re: Fwd: [LARTC] HTB shaping & borrowing info
2006-05-27 22:11 Fwd: [LARTC] HTB shaping & borrowing info Stefano Mainardi
2006-05-27 23:24 ` Andreas Klauer
@ 2006-05-28 0:04 ` Stefano Mainardi
2006-05-28 0:25 ` Andreas Klauer
` (9 subsequent siblings)
11 siblings, 0 replies; 13+ messages in thread
From: Stefano Mainardi @ 2006-05-28 0:04 UTC (permalink / raw)
To: lartc
[-- Attachment #1.1: Type: text/plain, Size: 824 bytes --]
2006/5/28, Andreas Klauer <Andreas.Klauer@metamorpher.de>:
>
> So, in other words, A is allowed to take bandwidth from B.
> B and C stick to their bandwidth limits.
There is a misunderstanding, sorry.
I need to know if it is possible to manage in a dynamic way the assignment
of traffic to workstations and of the traffic that they could use.
Like I said above, in the case that B is not producing traffic, 7/8 of the
20 MB/s need to be assigned to A and the remaining 1/8 will remain to B.
As final result A,B,C could exchange dynamically their "banda".
Is possible, to change dynamically their band assignment?
--
Stefano Mainardi
Presidente Associazione ILDN - Italian Linux Distro Network
Mobile: 349/3917212
Skype: mainardistefano
IM (ICQ): 250-292-408
Blog: http://www.mainardistefano.org
[-- Attachment #1.2: Type: text/html, Size: 1202 bytes --]
[-- Attachment #2: Type: text/plain, Size: 143 bytes --]
_______________________________________________
LARTC mailing list
LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
^ permalink raw reply [flat|nested] 13+ messages in thread* Re: Fwd: [LARTC] HTB shaping & borrowing info
2006-05-27 22:11 Fwd: [LARTC] HTB shaping & borrowing info Stefano Mainardi
2006-05-27 23:24 ` Andreas Klauer
2006-05-28 0:04 ` Stefano Mainardi
@ 2006-05-28 0:25 ` Andreas Klauer
2006-05-29 15:28 ` Stefano Mainardi
` (8 subsequent siblings)
11 siblings, 0 replies; 13+ messages in thread
From: Andreas Klauer @ 2006-05-28 0:25 UTC (permalink / raw)
To: lartc
On Sun, May 28, 2006 at 02:04:57AM +0200, Stefano Mainardi wrote:
> Like I said above, in the case that B is not producing traffic, 7/8 of the
> 20 MB/s need to be assigned to A and the remaining 1/8 will remain to B.
Well, reducing the ceil of A by 1/8 of B's bandwidth in the tree I
posted earlier would do that.
> Is possible, to change dynamically their band assignment?
The bandwidth in HTB is dynamic, as classes are allowed to borrow
bandwidth from other classes depending on their rate-ceil settings.
In the tree I posted, the bandwidth behaviour is as follows:
10mbit will be reserved for C at all times, B can use up to 20mbit,
A has 70mbit reserved, but can also use 20mbit of B if B is idle.
If the borrowing/lending bandwidth between HTB classes is not
dynamic enough for you, the only other option you have is to
somehow externally delete/create new HTB classes on the fly,
which is not a good solution in most situations.
Regards
Andreas Klauer
_______________________________________________
LARTC mailing list
LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
^ permalink raw reply [flat|nested] 13+ messages in thread* Re: Fwd: [LARTC] HTB shaping & borrowing info
2006-05-27 22:11 Fwd: [LARTC] HTB shaping & borrowing info Stefano Mainardi
` (2 preceding siblings ...)
2006-05-28 0:25 ` Andreas Klauer
@ 2006-05-29 15:28 ` Stefano Mainardi
2006-05-30 11:02 ` Stefano Mainardi
` (7 subsequent siblings)
11 siblings, 0 replies; 13+ messages in thread
From: Stefano Mainardi @ 2006-05-29 15:28 UTC (permalink / raw)
To: lartc
[-- Attachment #1.1: Type: text/plain, Size: 935 bytes --]
2006/5/28, Andreas Klauer <Andreas.Klauer@metamorpher.de>:
>
> > Is possible, to change dynamically their band assignment?
>
> The bandwidth in HTB is dynamic, as classes are allowed to borrow
> bandwidth from other classes depending on their rate-ceil settings.
Thanks for the precious tips.
I've seen that "ceil rate" with TC is a constant, i want to know if is
possible to make dynamic based on the load of the network?
> If the borrowing/lending bandwidth between HTB classes is not
> dynamic enough for you, the only other option you have is to
> somehow externally delete/create new HTB classes on the fly,
> which is not a good solution in most situations.
With this solution, i can think to create a series of HTB classes for
various situation.
But is impossible to think! :(
Or create a daemon that watch the load of the network and switch from HTB
classes to another.
--
Stefano Mainardi
[-- Attachment #1.2: Type: text/html, Size: 1372 bytes --]
[-- Attachment #2: Type: text/plain, Size: 143 bytes --]
_______________________________________________
LARTC mailing list
LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
^ permalink raw reply [flat|nested] 13+ messages in thread* Re: Fwd: [LARTC] HTB shaping & borrowing info
2006-05-27 22:11 Fwd: [LARTC] HTB shaping & borrowing info Stefano Mainardi
` (3 preceding siblings ...)
2006-05-29 15:28 ` Stefano Mainardi
@ 2006-05-30 11:02 ` Stefano Mainardi
2006-06-01 20:33 ` Andy Furniss
` (6 subsequent siblings)
11 siblings, 0 replies; 13+ messages in thread
From: Stefano Mainardi @ 2006-05-30 11:02 UTC (permalink / raw)
To: lartc
[-- Attachment #1.1: Type: text/plain, Size: 1188 bytes --]
Andreas,
if I want that A,B,C can borrow bandwidth from each other?
How i can the structure of HTB tree?
Stefano
2006/5/28, Andreas Klauer <Andreas.Klauer@metamorpher.de>:
>
> On Sun, May 28, 2006 at 02:04:57AM +0200, Stefano Mainardi wrote:
> > Like I said above, in the case that B is not producing traffic, 7/8 of
> the
> > 20 MB/s need to be assigned to A and the remaining 1/8 will remain to B.
>
> Well, reducing the ceil of A by 1/8 of B's bandwidth in the tree I
> posted earlier would do that.
>
> > Is possible, to change dynamically their band assignment?
>
> The bandwidth in HTB is dynamic, as classes are allowed to borrow
> bandwidth from other classes depending on their rate-ceil settings.
>
> In the tree I posted, the bandwidth behaviour is as follows:
>
> 10mbit will be reserved for C at all times, B can use up to 20mbit,
> A has 70mbit reserved, but can also use 20mbit of B if B is idle.
>
> If the borrowing/lending bandwidth between HTB classes is not
> dynamic enough for you, the only other option you have is to
> somehow externally delete/create new HTB classes on the fly,
> which is not a good solution in most situations.
>
> Regards
> Andreas Klauer
>
[-- Attachment #1.2: Type: text/html, Size: 1490 bytes --]
[-- Attachment #2: Type: text/plain, Size: 143 bytes --]
_______________________________________________
LARTC mailing list
LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
^ permalink raw reply [flat|nested] 13+ messages in thread* Re: Fwd: [LARTC] HTB shaping & borrowing info
2006-05-27 22:11 Fwd: [LARTC] HTB shaping & borrowing info Stefano Mainardi
` (4 preceding siblings ...)
2006-05-30 11:02 ` Stefano Mainardi
@ 2006-06-01 20:33 ` Andy Furniss
2006-06-01 20:37 ` Stefano Mainardi
` (5 subsequent siblings)
11 siblings, 0 replies; 13+ messages in thread
From: Andy Furniss @ 2006-06-01 20:33 UTC (permalink / raw)
To: lartc
Stefano Mainardi wrote:
> tc class add dev eth1 parent 11:1 classid 11:2 htb rate 70Mbit ceil 70Mbit
> burst 6k cburst 64k quantum 1600
I am not sure if it matters in this case with the big cburst - but
remember burst * 8 * HZ will be the bit rate a class can reach.
Andy.
_______________________________________________
LARTC mailing list
LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
^ permalink raw reply [flat|nested] 13+ messages in thread* Re: Fwd: [LARTC] HTB shaping & borrowing info
2006-05-27 22:11 Fwd: [LARTC] HTB shaping & borrowing info Stefano Mainardi
` (5 preceding siblings ...)
2006-06-01 20:33 ` Andy Furniss
@ 2006-06-01 20:37 ` Stefano Mainardi
2006-06-01 20:46 ` Stefano Mainardi
` (4 subsequent siblings)
11 siblings, 0 replies; 13+ messages in thread
From: Stefano Mainardi @ 2006-06-01 20:37 UTC (permalink / raw)
To: lartc
[-- Attachment #1.1: Type: text/plain, Size: 625 bytes --]
Hi Andy,
how i must setup the value for CBURST for my situation? Have you read?
Ste
2006/6/1, Andy Furniss <lists@andyfurniss.entadsl.com>:
>
> Stefano Mainardi wrote:
>
> > tc class add dev eth1 parent 11:1 classid 11:2 htb rate 70Mbit ceil
> 70Mbit
> > burst 6k cburst 64k quantum 1600
>
> I am not sure if it matters in this case with the big cburst - but
> remember burst * 8 * HZ will be the bit rate a class can reach.
>
> Andy.
>
--
Stefano Mainardi
Presidente Associazione ILDN - Italian Linux Distro Network
Mobile: 349/3917212
Skype: mainardistefano
IM (ICQ): 250-292-408
Blog: http://www.mainardistefano.org
[-- Attachment #1.2: Type: text/html, Size: 994 bytes --]
[-- Attachment #2: Type: text/plain, Size: 143 bytes --]
_______________________________________________
LARTC mailing list
LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
^ permalink raw reply [flat|nested] 13+ messages in thread* Re: Fwd: [LARTC] HTB shaping & borrowing info
2006-05-27 22:11 Fwd: [LARTC] HTB shaping & borrowing info Stefano Mainardi
` (6 preceding siblings ...)
2006-06-01 20:37 ` Stefano Mainardi
@ 2006-06-01 20:46 ` Stefano Mainardi
2006-06-01 20:46 ` Andy Furniss
` (3 subsequent siblings)
11 siblings, 0 replies; 13+ messages in thread
From: Stefano Mainardi @ 2006-06-01 20:46 UTC (permalink / raw)
To: lartc
[-- Attachment #1.1: Type: text/plain, Size: 957 bytes --]
Ok, i try without CBURST value.
Thanks ;)
2006/6/1, Andy Furniss <lists@andyfurniss.entadsl.com>:
>
> Stefano Mainardi wrote:
> > Hi Andy,
> > how i must setup the value for CBURST for my situation? Have you read?
> >
>
> I think the burst is too small - In your case I would just not specify
> either burst and let htb choose them.
>
> Andy.
>
> > Ste
> >
> > 2006/6/1, Andy Furniss <lists@andyfurniss.entadsl.com>:
> >
> >>
> >> Stefano Mainardi wrote:
> >>
> >> > tc class add dev eth1 parent 11:1 classid 11:2 htb rate 70Mbit ceil
> >> 70Mbit
> >> > burst 6k cburst 64k quantum 1600
> >>
> >> I am not sure if it matters in this case with the big cburst - but
> >> remember burst * 8 * HZ will be the bit rate a class can reach.
> >>
> >> Andy.
> >>
> >
> >
> >
>
>
--
Stefano Mainardi
Presidente Associazione ILDN - Italian Linux Distro Network
Mobile: 349/3917212
Skype: mainardistefano
IM (ICQ): 250-292-408
Blog: http://www.mainardistefano.org
[-- Attachment #1.2: Type: text/html, Size: 1496 bytes --]
[-- Attachment #2: Type: text/plain, Size: 143 bytes --]
_______________________________________________
LARTC mailing list
LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
^ permalink raw reply [flat|nested] 13+ messages in thread* Re: Fwd: [LARTC] HTB shaping & borrowing info
2006-05-27 22:11 Fwd: [LARTC] HTB shaping & borrowing info Stefano Mainardi
` (7 preceding siblings ...)
2006-06-01 20:46 ` Stefano Mainardi
@ 2006-06-01 20:46 ` Andy Furniss
2006-06-01 20:59 ` Andy Furniss
` (2 subsequent siblings)
11 siblings, 0 replies; 13+ messages in thread
From: Andy Furniss @ 2006-06-01 20:46 UTC (permalink / raw)
To: lartc
Stefano Mainardi wrote:
> Hi Andy,
> how i must setup the value for CBURST for my situation? Have you read?
>
I think the burst is too small - In your case I would just not specify
either burst and let htb choose them.
Andy.
> Ste
>
> 2006/6/1, Andy Furniss <lists@andyfurniss.entadsl.com>:
>
>>
>> Stefano Mainardi wrote:
>>
>> > tc class add dev eth1 parent 11:1 classid 11:2 htb rate 70Mbit ceil
>> 70Mbit
>> > burst 6k cburst 64k quantum 1600
>>
>> I am not sure if it matters in this case with the big cburst - but
>> remember burst * 8 * HZ will be the bit rate a class can reach.
>>
>> Andy.
>>
>
>
>
_______________________________________________
LARTC mailing list
LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
^ permalink raw reply [flat|nested] 13+ messages in thread* Re: Fwd: [LARTC] HTB shaping & borrowing info
2006-05-27 22:11 Fwd: [LARTC] HTB shaping & borrowing info Stefano Mainardi
` (8 preceding siblings ...)
2006-06-01 20:46 ` Andy Furniss
@ 2006-06-01 20:59 ` Andy Furniss
2006-06-01 21:13 ` Stefano Mainardi
2006-06-01 21:57 ` Andy Furniss
11 siblings, 0 replies; 13+ messages in thread
From: Andy Furniss @ 2006-06-01 20:59 UTC (permalink / raw)
To: lartc
Stefano Mainardi wrote:
> Ok, i try without CBURST value.
Without burst aswell ...
Andy.
_______________________________________________
LARTC mailing list
LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
^ permalink raw reply [flat|nested] 13+ messages in thread* Re: Fwd: [LARTC] HTB shaping & borrowing info
2006-05-27 22:11 Fwd: [LARTC] HTB shaping & borrowing info Stefano Mainardi
` (9 preceding siblings ...)
2006-06-01 20:59 ` Andy Furniss
@ 2006-06-01 21:13 ` Stefano Mainardi
2006-06-01 21:57 ` Andy Furniss
11 siblings, 0 replies; 13+ messages in thread
From: Stefano Mainardi @ 2006-06-01 21:13 UTC (permalink / raw)
To: lartc
[-- Attachment #1.1: Type: text/plain, Size: 408 bytes --]
Ops BURST...sorry!
And CBURST value you think that is correct?
2006/6/1, Andy Furniss <lists@andyfurniss.entadsl.com>:
>
> Stefano Mainardi wrote:
> > Ok, i try without CBURST value.
>
> Without burst aswell ...
>
> Andy.
>
--
Stefano Mainardi
Presidente Associazione ILDN - Italian Linux Distro Network
Mobile: 349/3917212
Skype: mainardistefano
IM (ICQ): 250-292-408
Blog: http://www.mainardistefano.org
[-- Attachment #1.2: Type: text/html, Size: 732 bytes --]
[-- Attachment #2: Type: text/plain, Size: 143 bytes --]
_______________________________________________
LARTC mailing list
LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
^ permalink raw reply [flat|nested] 13+ messages in thread* Re: Fwd: [LARTC] HTB shaping & borrowing info
2006-05-27 22:11 Fwd: [LARTC] HTB shaping & borrowing info Stefano Mainardi
` (10 preceding siblings ...)
2006-06-01 21:13 ` Stefano Mainardi
@ 2006-06-01 21:57 ` Andy Furniss
11 siblings, 0 replies; 13+ messages in thread
From: Andy Furniss @ 2006-06-01 21:57 UTC (permalink / raw)
To: lartc
Stefano Mainardi wrote:
> Ops BURST...sorry!
> And CBURST value you think that is correct?
I would leave that out aswell and let htb choose it from the rate of the
class.
Andy.
_______________________________________________
LARTC mailing list
LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
^ permalink raw reply [flat|nested] 13+ messages in thread