There has been a great deal of discussion on Quality of Service (QoS) and the tiering of the Internet.  In many peoples’ view, QoS technology will be used to create the multi-tier Internet.  Instead, I propose that QoS can’t really deliver on the multi-tier network. If it is thoughtlessly deployed it might actually break the existing economic and technical paradigms for broadband service.  Quite simply, mediation services must be used to manage the capacity of the network so that competing applications can receive appropriate levels of service. QoS is about reserving network capacity so that certain applications can receive “guaranteed” service.  The distinction sounds subtle but it is actually quite profound.

What really is QoS? Quality of Services refers to offering some level of guaranteed service to the applications running over the network.  The available network capacity can be quantified in many ways. Ordinarily, packet loss, delay, some amount of fixed bandwidth (bits per second), or when it comes to TCP a statistical description of congestion event measures (CEMs) and bandwidth delay product (BDP) can be used to describe that available “guaranteed” capacity. Thus QoS describes what the available network capacity is to specific applications as opposed to how that capacity is used by ordinary “best-effort” applications. While both technological approaches eliminate the last-mile bottleneck, only mediation technology allows the ISP to control how those bottlenecks allocate capacity to existing and new applications running over their infrastructure.

Thus, QoS is focused reserving bandwidth for one particular service and if “too much” bandwidth is reserved, then existing applications using the remaining best-effort capacity could break. Instead, we describe mediation as the process of discerning which broadband traffic application types are to obtain different levels of service. For example, using stateless mediation technology it is possible to offer existing best-effort applications like web surfing (e.g. HTTP) and email (e.g. POP3/SMTP/IMAP) with exactly the same service levels that they receive today while enabling new authorized streaming services (e.g. ISTP) to obtain any feasible bandwidth above a certain level (say 800Kbps). Under a QoS approach, the ISP would simply reserve the premium capacity for the streaming traffic while not attempting to control or measure the impact on the existing best-effort applications. This could be very dangerous to both the consumer and the broadband ISP if that provisioning was not carefully monitored and dynamically controlled.

Some make the claim that, at its root, that there would be no need for QoS (or mediation technology) if “cleanly” provisioned best-effort networks were put in place.  In other words, they claim that everyone can receive the level of service they require for their application by simply making the “highways wider”. This argument completely ignores the fact that there are applications out there like peer-to-peer (P2P) which effectively exploit the flat-rate cost best-effort system currently in place.  Current estimates indicate that on some networks up to 80% of the volume is now P2P traffic. With the Internet’s transition to distributing a video over traditional unicast methods as well as P2P, we can expect that this volume increase significantly.  Therefore, the “just make the highways” bigger argument completely ignores the fact that broadband service providers do not have the tools to adequately manage how their network capacity is used.  While QoS technology may not be the answer, mediation technologies most certainly can address the problem.