LARTC - Jun 2001 - Priotizing Bandwith but without Shaping the Bandwith

Hello, I''m Alexandra

Well, i have been working with cbq around 3 month ago without any
problems, what i do is control the bandwith to 64Kbit for input and
output and it works well.

But now I don''t want control the Bandwith, i only need prioritize udp
trafic over tcp trafic

Can somebody to help me with my requirement?


Thanks

Alexandra

Thanks for answer to me, but:

I have Red Hat 7.0 with kernel 2.2.16-22, and it doesn''t support
"mangle" options, how can i do the same "Priotizing udp traffic
over tcp
but without Shaping the Bandwith" with ipchains and cbq or only cbq?

Thanks

A.A.A.A.B.

"Marek L. Kozak" wrote:
> 
> Alexandra Alvarado wrote:
> 
> Hello,
> 
> > But now I don''t want control the Bandwith, i only need
prioritize udp
> > trafic over tcp trafic
> Try this:
> iptables -t mangle -A PREROUTING -t udp -j TOS --set-tos
> Maximize-Throughput
> --
> Regards,
> Marek L. Kozak

"Wingtung.Leung" wrote:
> 
> On Thu, 28 Jun 2001, Alexandra Alvarado wrote:
> 
> > I have Red Hat 7.0 with kernel 2.2.16-22, and it doesn''t
support
> > "mangle" options, how can i do the same "Priotizing udp
traffic over tcp
> > but without Shaping the Bandwith" with ipchains and cbq or only
cbq?
> >
> > "Marek L. Kozak" wrote:
> > >
> > > Try this:
> > > iptables -t mangle -A PREROUTING -t udp -j TOS --set-tos
> > > Maximize-Throughput
> 
> (Please quote correctly, and type your own text *beneath* the quoted text.
> It makes the message much more readable.)
> 
> First of all, I don''t think setting TOS field will change
behaviour of the
> network queue, most routers just ignore that field. And even if it is used
> on the local Linux router, I don''t think Alexandra wanted a high
> throughput for UDP traffic, but rather a low delay situation.
> 
> Please try to describe in detail what you mean with "giving priority
to
> UDP traffic over TCP traffic", and what you want to achieve.
> 
> The HOWTO gives an example for splitting bandwith with unbounded classes,
> it might be good enough for your situation. Please try it first, and if it
> is not what you wanted, give the problems with that setup.
> 
> I''m not sure, but I don''t think ipchains (kernel 2.2)
supports setting the
> TOS field. It doesn''t matter, you don''t really need a
firewall marker, you
> can also redirect the traffic using the general u32 filter.
> 
> --
> GnuPG public key: http://www.keyserver.net
> fingerprint = A3C4 DE50 712D 4FA8 C564  4D96 5E06 C9CC ECFA 19C5
> 
> _______________________________________________
> LARTC mailing list / LARTC@mailman.ds9a.nl
> http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO:
http://ds9a.nl/2.4Routing/

Hi, I''m Alexandra

Well, as i say before, a i need that all traffic udp has priority over
tcp traffic, because i have to improve the voice quility over internet,
well now i''m upgrading to Red Hat 7.1 (kernel 2.4) to use TOS field,
but
my question is if this sentence:

iptables -t mangle -A PREROUTING -p udp -j TOS --set-tos
Maximize-Throughput

works or not, and how can i verify it?

Until know i use CBQ, but with cbq i only asign specific bandwith to my
clients using ip address, but i don''t know how to prioritize traffic
udp
over tcp but without limiting the bandwith using CBQ.


Thanks

A.A.A.A.B.

On Thu, 28 Jun 2001, Alexandra Alvarado wrote:
> I have Red Hat 7.0 with kernel 2.2.16-22, and it doesn''t support
> "mangle" options, how can i do the same "Priotizing udp
traffic over tcp
> but without Shaping the Bandwith" with ipchains and cbq or only cbq?
>
> "Marek L. Kozak" wrote:
> >
> > Try this:
> > iptables -t mangle -A PREROUTING -t udp -j TOS --set-tos
> > Maximize-Throughput
(Please quote correctly, and type your own text *beneath* the quoted text.
It makes the message much more readable.)

First of all, I don''t think setting TOS field will change behaviour of
the
network queue, most routers just ignore that field. And even if it is used
on the local Linux router, I don''t think Alexandra wanted a high
throughput for UDP traffic, but rather a low delay situation.

Please try to describe in detail what you mean with "giving priority to
UDP traffic over TCP traffic", and what you want to achieve.

The HOWTO gives an example for splitting bandwith with unbounded classes,
it might be good enough for your situation. Please try it first, and if it
is not what you wanted, give the problems with that setup.

I''m not sure, but I don''t think ipchains (kernel 2.2) supports
setting the
TOS field. It doesn''t matter, you don''t really need a firewall
marker, you
can also redirect the traffic using the general u32 filter.



-- 
GnuPG public key: http://www.keyserver.net
fingerprint = A3C4 DE50 712D 4FA8 C564  4D96 5E06 C9CC ECFA 19C5

On Thu, 28 Jun 2001, Alexandra Alvarado wrote:
> Well, as i say before, a i need that all traffic udp has priority over
> tcp traffic, because i have to improve the voice quility over internet,
> well now i''m upgrading to Red Hat 7.1 (kernel 2.4) to use TOS
field, but
> my question is if this sentence:
>
> iptables -t mangle -A PREROUTING -p udp -j TOS --set-tos
> Maximize-Throughput
>
> works or not, and how can i verify it?
I''m not sure, but I *think* this does *not* work.
> Until know i use CBQ, but with cbq i only asign specific bandwith to my
> clients using ip address, but i don''t know how to prioritize
traffic udp
> over tcp but without limiting the bandwith using CBQ.
A simple solution: attach 2 unbounded (!) classes to your interface, both
with a rather low transfer rate like 20kbit/s and give one a high
priority and the other one low priority. There is a priority parameter
when you create a class.

Now find a method to redirect both UDP and TCP traffic to the correct
class. An easy and lazy way is to mark the packets with iptables in the
mangle table. For example, mark UDP packets with 1 and TCP packets with 2.
Look at the iptables manpage for info about using it.

Use tc and install the fw filters which redirect the marked packets into
the correct classes. Redirect the packets marked with 1 to the low hi prio
class, and marked with 2 to the low prio class.

That should to it.

The low transfer rate is the minimum garantued rate for that class, all
bandwidth which is not allocated to a class, will be delivered to the
class with the highest priority.

I described a similar setup a few weeks/months ago, maybe reading the
archives gives you more hints.

Succes.


(Good that you type beneath the quote, the second step is cut irrelevant
parts from the quoted text. Signatures for example. And this small note
too. :-)

-- 
GnuPG public key: http://www.keyserver.net
fingerprint = A3C4 DE50 712D 4FA8 C564  4D96 5E06 C9CC ECFA 19C5