Skip to contentSkip to conversation

HTB timer??

LARTC Lars Landmark 2 participants 4 messages May 14, 2003 – May 16, 2003

lore
Share & tools
lore
#0 lore raw

Hi;
As I understand, HTB use a timer in order to calculate rate. In addition it also use timer to tell whether classes are allowed to send packets or not.
Could somebody tell me how this timer is calculated??

If you have, say, a link with capacity 10MB. If you so configure using HTB two classes with the same parent to use 5MB each with priority 0,1 and ceil 10MB. What would happen if bandwidth decrease to under 5MB.

Would the highest prio class be the only served??

Thanks in advance

Lars
Student

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

#1 lore raw
Quoted 5 lines · Lars Landmark

Hi;
As I understand, HTB use a timer in order to calculate rate. In addition it also use timer to tell whether classes are allowed to send packets or not.
Could somebody tell me how this timer is calculated??

http://www.docum.org/stef.coene/qos/faq/cache/40.html

Quoted 5 lines

If you have, say, a link with capacity 10MB. If you so configure using HTB two classes with the same parent to use 5MB each with priority 0,1 and ceil 10MB. What would happen if bandwidth decrease to under 5MB.

Would the highest prio class be the only served??

No. Each class can send it's rate as a minimum. Prio is only used to distribute remaining bandwidth and to determine the order of the classes so the highest priority class has the lowest delays.
If you have a 10mbit class with 2 classes of each 2mbit and prio 0 and 1. Each class can send 2mbit. And the class with prio 0 can use the remaining 6mbit. So you have 2 classes with rate 2mbit and 8mbit.

Remember : the lower the prio, the higher the priority.

Stef

Signature · 8 lines
[email protected]
 "Using Linux as bandwidth manager"
     http://www.docum.org/
     #lartc @ irc.oftc.net

_______________________________________________
LARTC mailing list / [email protected]
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
#2 lore raw
In reply to Lars Landmark
Quoted 13 lines · Stef Coene
Quoted 5 lines · Lars Landmark

Hi;
As I understand, HTB use a timer in order to calculate rate. In addition it also use timer to tell whether classes are allowed to send packets or not.
Could somebody tell me how this timer is calculated??

http://www.docum.org/stef.coene/qos/faq/cache/40.html

Quoted 5 lines

If you have, say, a link with capacity 10MB. If you so configure using HTB two classes with the same parent to use 5MB each with priority 0,1 and ceil 10MB. What would happen if bandwidth decrease to under 5MB.

Would the highest prio class be the only served??

Quoted 8 lines

No. Each class can send it's rate as a minimum. Prio is only used to distribute remaining bandwidth and to determine the order of the classes so the highest priority class has the lowest delays.
If you have a 10mbit class with 2 classes of each 2mbit and prio 0 and 1. Each class can send 2mbit. And the class with prio 0 can use the remaining 6mbit. So you have 2 classes with rate 2mbit and 8mbit.

Remember : the lower the prio, the higher the priority.

Ok, But as you say prio is used in order:
to to determine the order of the classes so the highest priority class has the lowest delays.

As I understand this, high prio class must be placed among the first classes in the self feed list.
So my question is, how does HTB schedule this??? Among leafs classes it has a self feed list for each priority.
Does it place high prio class at first place, or how does it order this list.??
I have done a test for this scenario, but the result as you say did not end with starving low prio class.

Lars
Student

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

#3 lore raw
In reply to Lars Landmark
Quoted 6 lines
Quoted 1 line

Remember : the lower the prio, the higher the priority.

Ok, But as you say prio is used in order:
to to determine the order of the classes so the highest priority class has the lowest delays.

Quoted 8 lines

As I understand this, high prio class must be placed among the first classes in the self feed list.
So my question is, how does HTB schedule this??? Among leafs classes it has a self feed list for each priority.
Does it place high prio class at first place, or how does it order this list.??
I have done a test for this scenario, but the result as you say did not end with starving low prio class.

It also depends on how much data you generate in the low prio class. If you generate more data then the rate, the latency will be very high. You can find this on the htb homepage.
So for high latency, never send more data in the class then it's configured rate.

Stef

Signature · 8 lines
[email protected]
 "Using Linux as bandwidth manager"
     http://www.docum.org/
     #lartc @ irc.oftc.net

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