[LARTC] Quality of Service with VLAN using Traffic Control (tc)

[Johnray Fuller <jrfuller@redhat.com>]

We are using VLAN tagging to put several networks on one machine.

To ensure quality of service on one of the VLANs, we would like to
prioritize (outgoing/egress traffic) one VLAN interface over the other
(or others), but without capping bandwidth on any of the interfaces.
We want to do this while being careful to minimize increases in
latency on the prioritized interface.

Since these are two or more separate VLAN interfaces on the same
physical NIC, we want to know if specific policies assigned to each
interface through the tc command would be best suited for this task. At
the moment we are using the classless prio_fifo qdisc parameter in
tc. Are there other qdisciplines we should consider (such as the 
classful HTB). In other words, what tc policies have the best low 
latency principles for the prioritized VLAN?

Also under consideration are IMQ meta-interfaces for system wide
traffic shaping across multiple interfaces. But does  the use of IMQ and 
the requisite iptables packet tagging have a siginificant impact on 
increases latency in your experience?

Finally, is a weighted round-robin queuing algorithm supported. I was 
unable to figure this out.

Thanks for any assistance you can provide.

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