March 18th, 2005

VoIP QoS Bullshit


I don’t believe in Quality-of-Service (QoS). I don’t buy all the vendors FUDs about “ need QoS or else the world is going to end so buy my this special QoS-enabled box! now!”.

It is not that I don’t believe in quality for voice – I do – but I don’t think we need special QoS technology – 802.11e or whatever – to get there.

So when I pointed this out in the Wifi SIP Summit, someone from the audience quickly debunk me – and later pointed out the “QoS study” by ClearVoice. Nevermind I actually did it at APRICOT with roaring reviews – “no no no, you need QoS!” Ha!Now, I think the QoS study is pretty right on the mark : you can probably support 6-12 concurrent calls on one base station and really, I don’t need any study to tell me that. Get a base station and do some large file transfer concurrent and watch the throughput – Normal ones gives you about 500-600kbps and good one gives you 1.2mbps max then divide this by G.711 64kbps x 2 = 6-12 concurrent calls.

This is a limitation of physic – QoS can’t increase the number of calls. The best it can do is to prioritize the RTP.

But it still does not matter. Here are two reasons why:

1. Utilization for phone calls are typically 1-2% => you can support 300-1200 people with 6-12 concurrent calls. But lets just say utilization is higher for WiFi say 10%, you still have 60-120 people.

Now, find me a sane Wifi network admin who will put 60-120 people on *one* base station. No, you can’t – you just cant have 60-120 users on wifi network on one base station. You need at 3-4 base stations to support that many users.

So the argument 6-12 concurrent calls on one base station is really moot – the capacity is more then the utilization at least by today calculation.

2. Codecs is improving which reduced bandwidth per call session down to 16kbps or even 8kbps from 64kbps. Then coupled with ever increasing bandwidth (thanks to better wireless technology) from 54mbps (11a) and >100mbps (promised by 11n), bandwidth scarity is also moot.

Theortically, using G.729, you can support 8x more (ie 48) concurrent calls on 11b and 20x more on 11a (ie 120). You going to have more problem with other apps (ie. streaming) your users going to use then VoIP.

QoS for Wifi like 802.11e sound exactly like RSVP for landline – sound good in theory (your leaseline is expensive so you better prioritize your traffic), good enough to sell boxes but no one uses it much in the long run. So don’t listen to vendors bullshit.

ps: BTW, indoors GSM base station can only handle 6 concurrent calls within 500m so 6-12 concurrent calls on indoor wifi is pretty good by cost-benefit analysis.

Comments are closed.