Skip to contentSkip to conversation

Qos Slowing down

LARTC Thomas Switala 2 participants 2 messages Nov 19, 2003

lore
Share & tools
lore
#0 lore raw

Hi

I have a bandwidth manager configured on the following system:

RedHat 8.0
Kernel - 2.4.20 (Recompiled and I changed the
SFQ_DEPTH to 512
and the SFQ_HASH 3072 and the PSCHED_CLOCK_SOURCE
= PSCHED_CPU

Patches for the kernel = ebtables glibc-2.3.2-4.80.6

The system is build to act as a bridge. Qos happens on the outgoing eth0 device to about
175 differant subnets. I am using HTB and I have the following hierachy

                                                Group1
                                                    |
                -----------------------------------------
                |
|
            SP1                                                        SP2
                |
|
-------------                                              -------------
 |      |     |    |                                               |     |
|     |
cl1 cl2 cl3 cl4                                            cl5 cl6  cl7  cl8

Al of the cl belong to a service provider (sp) where they are on shared bandwidth
all of them have a CIR of 1 and a Ceil of 384kbps.

The problem that I am experiencing , and it happens often lately is that the Qos slows down and
it doesn't pass any traffic at all. I have to restart the whole system before it starts working again.
If I stop the tc on the device it works as well which rules out the bridging.

What could be wrong and is Linux going to handle more than 800 sites on a 40Mbps connection ?

--
Warm Regards
Thomas I. Switala
CEO and Lead Systems Architect Genotribe Ltd

--
Cell : +27-82-822-5080 (South-Africa)

            +234-803-794-8252   (Nigeria)

IM: [email protected] (Jabber)
genotribe (Yahoo)

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

#1 lore raw
Quoted 46 lines · Thomas Switala

Hi

I have a bandwidth manager configured on the following system:

RedHat 8.0
Kernel - 2.4.20 (Recompiled and I changed the
SFQ_DEPTH to 512
and the SFQ_HASH 3072 and the PSCHED_CLOCK_SOURCE
= PSCHED_CPU

Patches for the kernel = ebtables glibc-2.3.2-4.80.6

The system is build to act as a bridge. Qos happens on the outgoing eth0 device to about
175 differant subnets. I am using HTB and I have the following hierachy

                                                Group1

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


            SP1                                                        SP2

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

cl1 cl2 cl3 cl4 cl5 cl6 cl7
cl8

Al of the cl belong to a service provider (sp) where they are on shared bandwidth
all of them have a CIR of 1 and a Ceil of 384kbps.

The problem that I am experiencing , and it happens often lately is that the Qos slows down and
it doesn't pass any traffic at all. I have to restart the whole system before it starts working again.
If I stop the tc on the device it works as well which rules out the bridging.

Quoted 2 lines

What could be wrong and is Linux going to handle more than 800 sites on a 40Mbps connection ?

Do you see any information in the kernel log? And what shows dmesg? Maybe a spu or memory bottleneck?

Stef

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

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