Interest in quality of service in IP networks is growing, but its role and value remain contentious. Three types of issue are contributing to the debate.
First is the mildly confusing acronym “QOS” which can mean either “quality of service” or “mechanism for achieving a desired quality of service”; second is a range of QOS-related differences between public and private networks — and between WANs and LANs; and third, of course, is politics. Let’s reduce the challenge in addressing these issues by considering the example of network congestion, one of the most important QOS concerns.
Solving certain congestion problems can be achieved by adding capacity or implementing some form of traffic prioritization. Using the first definition of “QOS” given above, if you resort to either solution you are seeking improved QOS. Using the second definition, only traffic prioritization is considered to be QOS; therefore, if you add capacity, you are not employing QOS but you are still seeking improved quality of service. While some may see this as semantic trivia, others are confused by the dual meaning. Below, I adopt the second definition (i.e., a “mechanism…”) for convenience.
The first motivation for using QOS is technical: certain types of traffic (e.g., voice) are less tolerant of the effects of congestion than other types (e.g., email). Therefore, it is valid to consider giving priority to less tolerant types of traffic for technical reasons. The second motivation is market-based and applies mainly to public network services: certain customers may be prepared to pay more for higher quality services. This case is not so clear cut, since there are those who see this type of prioritization as a threat to the neutrality of public networks, thus introducing a political dimension to the issue.
A common argument against QOS is that adding more capacity may be easier and less costly. But that is subject to circumstances; for example, adding capacity to an international network would have much different economics than adding capacity to local services. It has been claimed that QOS should be a last resort, on the basis that it deliberately degrades overall network performance in order to improve the quality for a minority of the applications. But this is misleading.
Firstly, it relies on a definition of network performance that excludes quality as a factor; there is not much point in claiming a higher throughput, if some of the content is unintelligible. Secondly, it ignores the fact that certain applications can easily tolerate significant delays; therefore, to talk about degrading those applications in that particular sense is practically meaningless. Finally, it ignores the following consideration.
A third motivation for using QOS is importance-based and applies mainly to enterprise networks and emergency services on public networks. Certain users, terminals or applications may be more important than others, which is not taken into account when arguing that more capacity is the preferred solution. Under normal conditions a particular network capacity may be sufficient. However, under abnormal conditions, prioritization could provide a significant advantage to those who really warrant it. It is interesting to note that our discussion has shifted from quality of service to priority of service. POS is a different concept from QOS, and its increasing importance will make the QOS debate much more interesting.