All of lore.kernel.org
 help / color / mirror / Atom feed
* [LARTC] exact rate
@ 2005-12-14 13:10 Anton Glinkov
  2005-12-22 20:22 ` Jody Shumaker
                   ` (2 more replies)
  0 siblings, 3 replies; 4+ messages in thread
From: Anton Glinkov @ 2005-12-14 13:10 UTC (permalink / raw)
  To: lartc

Hello
I am using htb (krenel 2.6.14.3 + esfq patch from fatooh.org) to limit the
bandwidth. But I want the absolute maximum rate to be for example 5Mbit
and not to exceed it.

here is the script:

$tc qdisc add dev ${NETIF} root handle 1 htb r2q 4
$tc class add dev ${NETIF} parent 1: classid 1:21 htb rate 5Mbit ceil 5Mbit
$tc class add dev ${NETIF} parent 1:21 classid 1:101 htb rate 2Mbit ceil
5Mbit prio 1
$tc class add dev ${NETIF} parent 1:21 classid 1:102 htb rate 1Mbit ceil
5Mbit prio 2
$tc class add dev ${NETIF} parent 1:21 classid 1:103 htb rate 3Mbit ceil
5Mbit prio 3

when I don't specify burst the flow always exceeds 5Mbit
when I try to play with it (the burst) for example turning it off completely
burst 0 cburst 0 for parent and children classes the rate falls ot 2-3 Mbits

removing the esfq qdiscs doesn't change anything...

I want exact rate of 5Mbit for the parent class and no more?

any suggestions are welcome :)

-- 
Anton Glinkov
network administrator

_______________________________________________
LARTC mailing list
LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [LARTC] exact rate
  2005-12-14 13:10 [LARTC] exact rate Anton Glinkov
@ 2005-12-22 20:22 ` Jody Shumaker
  2005-12-23  2:08 ` Salim
  2005-12-23  5:35 ` Jody Shumaker
  2 siblings, 0 replies; 4+ messages in thread
From: Jody Shumaker @ 2005-12-22 20:22 UTC (permalink / raw)
  To: lartc


[-- Attachment #1.1: Type: text/plain, Size: 1850 bytes --]

Your subclass's combined rate's should never exceed your parent class's
rate, and in this case your parent class is 5mbit, yet your subclasses have
a combined guarenteed bandwidth of 6mbit.  In my setup,  the root classes
rate is followed strictly.   With burst, it may go over temporarily but over
say a minute's time, the bandwidth is pretty close to the set rate, usualy
within ~0.4%. I'm unsure on what the behavior is when you allocate more
bandwidth to subclasses than the root, but it may be the reason you're not
seeing accurate results.

- Jody

On 12/14/05, Anton Glinkov <lartc@ssi.bg> wrote:
>
> Hello
> I am using htb (krenel 2.6.14.3 + esfq patch from fatooh.org) to limit the
> bandwidth. But I want the absolute maximum rate to be for example 5Mbit
> and not to exceed it.
>
> here is the script:
>
> $tc qdisc add dev ${NETIF} root handle 1 htb r2q 4
> $tc class add dev ${NETIF} parent 1: classid 1:21 htb rate 5Mbit ceil
> 5Mbit
> $tc class add dev ${NETIF} parent 1:21 classid 1:101 htb rate 2Mbit ceil
> 5Mbit prio 1
> $tc class add dev ${NETIF} parent 1:21 classid 1:102 htb rate 1Mbit ceil
> 5Mbit prio 2
> $tc class add dev ${NETIF} parent 1:21 classid 1:103 htb rate 3Mbit ceil
> 5Mbit prio 3
>
> when I don't specify burst the flow always exceeds 5Mbit
> when I try to play with it (the burst) for example turning it off
> completely
> burst 0 cburst 0 for parent and children classes the rate falls ot 2-3
> Mbits
>
> removing the esfq qdiscs doesn't change anything...
>
> I want exact rate of 5Mbit for the parent class and no more?
>
> any suggestions are welcome :)
>
> --
> Anton Glinkov
> network administrator
>
> _______________________________________________
> LARTC mailing list
> LARTC@mailman.ds9a.nl
> http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
>

[-- Attachment #1.2: Type: text/html, Size: 2313 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] 4+ messages in thread

* Re: [LARTC] exact rate
  2005-12-14 13:10 [LARTC] exact rate Anton Glinkov
  2005-12-22 20:22 ` Jody Shumaker
@ 2005-12-23  2:08 ` Salim
  2005-12-23  5:35 ` Jody Shumaker
  2 siblings, 0 replies; 4+ messages in thread
From: Salim @ 2005-12-23  2:08 UTC (permalink / raw)
  To: lartc


[-- Attachment #1.1: Type: text/plain, Size: 2459 bytes --]

What confuses me is, how the popular HTB script 'wondershaper' works! The sum of its subclasses' rates is way over the parent's rate.
  ----- Original Message ----- 
  From: Jody Shumaker 
  To: Anton Glinkov 
  Cc: lartc@mailman.ds9a.nl 
  Sent: Friday, December 23, 2005 4:22 AM
  Subject: Re: [LARTC] exact rate


  Your subclass's combined rate's should never exceed your parent class's rate, and in this case your parent class is 5mbit, yet your subclasses have a combined guarenteed bandwidth of 6mbit.  In my setup,  the root classes rate is followed strictly.   With burst, it may go over temporarily but over say a minute's time, the bandwidth is pretty close to the set rate, usualy within ~0.4%. I'm unsure on what the behavior is when you allocate more bandwidth to subclasses than the root, but it may be the reason you're not seeing accurate results. 

  - Jody


  On 12/14/05, Anton Glinkov <lartc@ssi.bg> wrote:
    Hello
    I am using htb (krenel 2.6.14.3 + esfq patch from fatooh.org) to limit the
    bandwidth. But I want the absolute maximum rate to be for example 5Mbit
    and not to exceed it.

    here is the script:

    $tc qdisc add dev ${NETIF} root handle 1 htb r2q 4
    $tc class add dev ${NETIF} parent 1: classid 1:21 htb rate 5Mbit ceil 5Mbit
    $tc class add dev ${NETIF} parent 1:21 classid 1:101 htb rate 2Mbit ceil 
    5Mbit prio 1
    $tc class add dev ${NETIF} parent 1:21 classid 1:102 htb rate 1Mbit ceil
    5Mbit prio 2
    $tc class add dev ${NETIF} parent 1:21 classid 1:103 htb rate 3Mbit ceil
    5Mbit prio 3

    when I don't specify burst the flow always exceeds 5Mbit 
    when I try to play with it (the burst) for example turning it off completely
    burst 0 cburst 0 for parent and children classes the rate falls ot 2-3 Mbits

    removing the esfq qdiscs doesn't change anything...

    I want exact rate of 5Mbit for the parent class and no more?

    any suggestions are welcome :)

    --
    Anton Glinkov
    network administrator

    _______________________________________________
    LARTC mailing list 
    LARTC@mailman.ds9a.nl
    http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc





------------------------------------------------------------------------------


  _______________________________________________
  LARTC mailing list
  LARTC@mailman.ds9a.nl
  http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc

[-- Attachment #1.2: Type: text/html, Size: 3966 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] 4+ messages in thread

* Re: [LARTC] exact rate
  2005-12-14 13:10 [LARTC] exact rate Anton Glinkov
  2005-12-22 20:22 ` Jody Shumaker
  2005-12-23  2:08 ` Salim
@ 2005-12-23  5:35 ` Jody Shumaker
  2 siblings, 0 replies; 4+ messages in thread
From: Jody Shumaker @ 2005-12-23  5:35 UTC (permalink / raw)
  To: lartc

Thing is, it did not work for me, because of that exact reason.  I had 
issues with it ignoring rates and in general not giving the lower 
latency I was going for. I really wish someone would update it.  
lartc.org is broken right now, but I believe the version up there is 3 
years old.

- Jody

Salim wrote:

> What confuses me is, how the popular HTB script 'wondershaper' works! 
> The sum of its subclasses' rates is way over the parent's rate.
>
>     ----- Original Message -----
>     *From:* Jody Shumaker <mailto:jody.shumaker@gmail.com>
>     *To:* Anton Glinkov <mailto:lartc@ssi.bg>
>     *Cc:* lartc@mailman.ds9a.nl <mailto:lartc@mailman.ds9a.nl>
>     *Sent:* Friday, December 23, 2005 4:22 AM
>     *Subject:* Re: [LARTC] exact rate
>
>     Your subclass's combined rate's should never exceed your parent
>     class's rate, and in this case your parent class is 5mbit, yet
>     your subclasses have a combined guarenteed bandwidth of 6mbit.  In
>     my setup,  the root classes rate is followed strictly.   With
>     burst, it may go over temporarily but over say a minute's time,
>     the bandwidth is pretty close to the set rate, usualy within
>     ~0.4%. I'm unsure on what the behavior is when you allocate more
>     bandwidth to subclasses than the root, but it may be the reason
>     you're not seeing accurate results.
>
>     - Jody
>
>     On 12/14/05, *Anton Glinkov* <lartc@ssi.bg <mailto:lartc@ssi.bg>>
>     wrote:
>
>         Hello
>         I am using htb (krenel 2.6.14.3 <http://2.6.14.3> + esfq patch
>         from fatooh.org <http://fatooh.org>) to limit the
>         bandwidth. But I want the absolute maximum rate to be for
>         example 5Mbit
>         and not to exceed it.
>
>         here is the script:
>
>         $tc qdisc add dev ${NETIF} root handle 1 htb r2q 4
>         $tc class add dev ${NETIF} parent 1: classid 1:21 htb rate
>         5Mbit ceil 5Mbit
>         $tc class add dev ${NETIF} parent 1:21 classid 1:101 htb rate
>         2Mbit ceil
>         5Mbit prio 1
>         $tc class add dev ${NETIF} parent 1:21 classid 1:102 htb rate
>         1Mbit ceil
>         5Mbit prio 2
>         $tc class add dev ${NETIF} parent 1:21 classid 1:103 htb rate
>         3Mbit ceil
>         5Mbit prio 3
>
>         when I don't specify burst the flow always exceeds 5Mbit
>         when I try to play with it (the burst) for example turning it
>         off completely
>         burst 0 cburst 0 for parent and children classes the rate
>         falls ot 2-3 Mbits
>
>         removing the esfq qdiscs doesn't change anything...
>
>         I want exact rate of 5Mbit for the parent class and no more?
>
>         any suggestions are welcome :)
>
>         --
>         Anton Glinkov
>         network administrator
>

_______________________________________________
LARTC mailing list
LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2005-12-23  5:35 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2005-12-14 13:10 [LARTC] exact rate Anton Glinkov
2005-12-22 20:22 ` Jody Shumaker
2005-12-23  2:08 ` Salim
2005-12-23  5:35 ` Jody Shumaker

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.