Quality Time on the Internet

By: Jimmy McGibney

Imagine the Waterford to Dublin train is free all the time. Four trains a day, five carriages on each train, but no ticket office, no booking system, and no restriction on the number of passengers that can get on the train. It's easy to manage for Iarnroid Eireann as there's no longer a need for conductors and ticket desks. It works well at certain quiet times, like Sunday mornings. Even at busier times, the train is quite comfortable on certain sections of the route, like the Waterford to Thomastown segment and maybe even on to Kilkenny, but gets congested and uncomfortable after that. One day, however, there's a big match in Croke Park, and the normally quiet Sunday morning train is packed for the entire journey. Even on "normal" days, it's impossible to predict in advance how busy the train is going to be as a few school tours could triple the number of people travelling.

The situation described above is very similar to today's Internet. People do generally have to pay to get on the Internet, but this is like paying for the petrol or bus fare to get you to the train station. After paying for access (generally your phone call to a local Internet service provider), everything else is free - it doesn't matter if you're emailing your friend up the street or someone in Japan, nor how big the email is.

The traffic on the Internet, just like on the train, is unpredictable. Sometimes it's quiet and you experience fast response times, and other times it's congested and you're thinking of the other interpretation of the WWW acronym - the World Wide Wait! Delays, errors and so on, may not be so serious if you're sending email or even surfing the web. However, if you're trying to use the Internet for real-time activities, like listening to music or having a telephone conversation, congestion can make it extremely difficult. The service provided by the Internet, often called "best-effort", tries to do what you require as efficiently as possible, but provides no guarantees.

So how do we address this problem? Part of the solution is quality of service, and work is ongoing in the Internet technology community on techniques to provide this. From the point of view of the communications industry, there is a major incentive for providing services with quality guarantees - users might be willing to pay for these services. People are used to getting Internet services for free, as it's unreasonable to expect anyone to pay for so-called "best effort" services - you're probably not going to pay for transferring a file from A to B if you're not fairly certain that it's going to get there in reasonable time.
Let's look again at our train analogy. One way to improve the passenger's experience is to provide different classes of service. Of course if it's still free or if all tickets cost the same, everyone will try to travel first class and we're no further on - so the solution is to charge more for first class. Having a first class ticket does not guarantee a seat, but at least there is a better ratio of seats to typical demand, and thus a better chance of getting one. An alternative approach to improving quality is to allow passengers to make a seat reservation, again at a cost. Now the passenger traveling from Dublin to Waterford knows that they're assured of a seat for the full journey, as does the passenger who gets on at Kildare. People are still allowed to buy (possibly cheaper) tickets without a reservation but they have no guarantees. On quiet days, it doesn't make much difference but the day lots of school tours take place they feel they're getting good value for money.

These two approaches, known as prioritisation and resource reservation, are precisely those being taken by the Internet standards community for managing data traffic so that users will be able to have a consistent experience of service quality, no matter when they connect. The TSSG at WIT are very active in this community, through several research projects.

Quality guarantees for the Internet are on the way. The only problem is that it'll probably cost us. Hopefully we'll get value for money!