QoS that works…

… Is what you get with the Tomato firmware for the Linksys WRT range of routers.

I have been messing around with DD-WRT trying to get it up and running in a workable state for VoIP, and my other $network->internet(); usage.

Unfortunately, everywhere I turned in DD-WRT for QoS lead me to issues after my initial setup didn’t work as expected at providing a shaping mechanism to slow the data rate of network services down, as well, as provide bandwidth for VoIP calls.

I began to think about how best to approach the issue a few weekends back, and came to a conclusion that the DD-WRT firmware needs to be pulled apart, and rebuilt with shaping that suits my network usage, mainly putting a set limit on all network traffic to 192kb, and placing services I want to be interactive with, like SSH, and VoIP, to be able to use all bandwidth, thus creating a “always available” buffer for them and ensuring that we don’t shape the network to hard to a state that uploading becomes unbearable.

Anyway, I read about the tomato firmware being good for QoS a while back, but didn’t want to take the time to tear down the DD-WRT config, and reload the Tomato configuration.

.. Until tonight, when I was talking with a MSN contact, and it came back up in conversation. I thought, stuff waiting, lets get it loaded up and test it all out.

And, I did, and it is very impressive in its reporting facilities.

I don’t want to place DD-WRT in the bin just yet, because I was looking at things differently. The general idea should be, you place all traffic in a pool, and pull out traffic you want to have running fast.

This router does that, but with the added bonus, you can configure the speeds that tc shapes at, and the iptables setup is a lot better.

The router as a whole seems great. But. And a big but. SNMP is gone!

I’ve found a workaround by holding on to SNMP in the CIFS dir, and loading it from there, it seems to work, but MRTG is broken. Good thing they included rstats :).

I’m thinking there’s other reasons for MRTG being broken, considering I followed the external program method to get the data for MRTG and provide it, and it still doesn’t seem to be taking it on board and updating my RRD and graphs.

That aside, the QoS benefits will outweigh the graphing benefits, and if it still doesn’t want to play ball, I can work around it by just going straight to rrdtool instead.

In other news, the routes for OzVoIPStatus seem to be fixed, it looks as though it was simply routed around, rather than corrected by WCG, so the original problem still exists, but just not for me, and TSN Internet customers.

In completely unrelated news, The G9 claim they don’t need Telstra’s business on their network, as they have enough business between them all to make it all viable, which is great news for them.

Even greater news is that the proposal is expected to finally be in the public eye this week, and I’m very keen to tear it apart! 🙂