Hi @all,

I would like to use the HFSC scheduling algorithm (because it produces good QoS results for VoIP). Unfortunately it is not included in kmod-sched.ipk. What steps do I have to do to get it running ?

I have a WRT54GS v. 1.1 so I am running the experimental release.
Many thanks in advance!

regards,
René

it is activated in cvs since 5 days.

You need to build on your own.

There is a bug in the kernel fixed, too. (related to tc)
So better use cvs or my snapshots:
http://openwrt.inf.fh-brs.de/~wbx/new

First thanks for your answer.

Does activated in cvs mean that I can just use it if I install the firmware from your snapshots ? Do I need additionally the kmod-sched ?
I am not sure now if I have to build on my own or just install one of your snapshots.

René

Hi there!

I'm using Kamikaze 2.6 on Asus wl-500gP and lately, using hfsc scheduler I encountered a strange behaviour....

It is shaping traffic *way* beyond the set criteria. For example, setting the desired bandwidth to 10000 kbit/s results in throughput arround 75kbit/sec but when rising the level to 11000+ Kbit/s the result is arround 1,2 Mbit/sec.
Lowering or rising the desired throughput dows not produce any "logical" results (at least for me). (ex. 5000 Kbit/s produces 65Kbit/sec, 1000Kbit/s 60Kbit/s, 50000Kbit/s 1,3Mbit/s...)

WBX, is the bug you mention related to this issue?

I tried the *same* script on Kamikaze 2.6 x86 and it is working flawlessly. Other schedulers (htb, cbq) are working as expected it's only the hfsc that gives such strange results....

Lurking through the code of sch_hfsc I noticed a "bug", which is clearly noted, regarding strange results on Gigabit speeds (did not tried to re-enable the disabled code).
I suppose the problem is somehow related to HZ (brcm: HZ=1000, x86: HZ=100) so I tried recompiling the schedulers using gettimeofday() but the results were the same....

Any ideas anyone?

Kind regards,
M.Culibrk