|
|
article continued from page 1
The result of this intervention is to redirect the focus of QoS for voice calls. Instead of being concerned primarily about the voice quality delivered, we can instead look at the proportion of attempted sessions that are successfully initiated. By avoiding network overload, we can replace the assumption that all sessions are successfully set up, with the assumption that all successful calls will receive acceptable voice quality.
From the customers’ perspective this changes the experience. Every successful call will sound good, but a few call attempts may be unsuccessful, and the customer will have to try again later. By comparison, with packet-level QoS the customer will find that almost all calls are connected “successfully”, but on some proportion of connected calls, network congestion will cause sound quality to be less than satisfactory.
|
|
"...what measurement is more useful: call failure rates, or some assessment of distortion and customer acceptability." |
|
the service providers’ perspective, call completion statistics provide hard information to assist capacity planning. Subjective opinions on voice call quality are perhaps not so helpful.
The QoS feature of IMS is therefore of fundamental importance. Carriers need session control to manage and deliver QoS in a way that is remarkably similar to the way it happens in the traditional PSTN network. IMS session control provides carriers with exactly the same useful parameter (call connection success rates) that they have used for years..
|
|
|
|
|
Which do customers prefer? I’m not sure anyone has asked them recently, at least not in a large-scale study. However, based on my unscientific observations I think the answer is, predictably: it depends. Business people in general, like good quality audio, but if the phone call is the one that’s going to clinch a deal seconds before the deadline, even a CEO with high standards will put up with some distortion, provided the message is understood. On the other hand, people who rely on cheap and cheerful VoIP to make international calls will generally be happy with lo-fi connections as long as the price is low. But if one call in three becomes so distorted that the call is useless, then those calls effectively fail anyway. Getting a busy tone might be less frustrating.
Note that if there is sufficient capacity to carry all sessions demanded by customers, it doesn’t really matter whether we adopt a packet-based or session-based approach. The calls will mostly be fine. So the QoS debate resolves to what measurement is more useful: call failure rates, or some assessment of distortion and customer acceptability. From
|
|
Are there alternatives to the IMS approach to quality management? Possibly. Throwing capacity at the problem before there is a problem is one way. Reducing bandwidth requirements of video and audio sessions would also help. It is also possible to improve the ability of end devices to measure audio/video performance and terminate the call quickly if a good session can’t be created. We can certainly expect to see continued developments in edge-device intelligence. It may be that ultimately cheap enough and good enough will be sufficient for commoditized voice/video sessions. No fancy network intelligence may be needed to deliver those benefits.
Meantime service providers need to control, measure, and manage to deliver services as profitably as possible. IMS, for the present, provides the network intelligence to do that.
|
|
|
|
