=There is considerable confusion in the consumer network access industry regarding the objective of congestion relief (traffic management) and that of market-segmentation and value-definition (usage management). In particular, many players often try to message that they are addressing congestion using volume limits, when in fact this is less effective.

I’ll refer to congestion occurring closest to the subscriber as the “weak link”.  Not surprisingly, the most effective (in terms of relieving the congestion) and most fair (to subscribers) way to manage this congestion is by implementing corrective policies that are localized at the site of the congestion itself. In fact, this precise approach is what the FCC defines as “narrowly tailored”.  Referring to Figure 1 below, if congestion occurs at the weak link (see A*) location and affects only users 1-5, then correcting that instance of congestion is ideal because Internet user traffic in other network locations such as locations B and C need not be touched.

Sandvine manages link congestion by examining, identifying, and prioritizing time-sensitive gaming, VoIP and streaming applications, over non time-sensitive applications such as emailing and browsing, or, alternatively, prioritising the users who are causing the least congestion vs. those who are causing the most in a short time interval.  You and I don’t notice if we get an email 2 or 3 seconds later, but we will be upset if our voice calls are garbled and we can’t understand each other. 

Managing congestion to the edge of the network, as close to the affected subscribers as possible, is a competitive differentiator for Sandvine because our competitors only help ISPs manage aggregate traffic patterns.  This aggregate approach is based on the flawed assumption that the macroscopic view (Figure 2) of Internet traffic activity is true for points A, B and C of the networks.  At Sandvine, we manage traffic to the edge of the network (Figure 3) and monitor points A, B and C individually.  This yields a more efficient network: higher utilisation, happier consumers, vs. the alternative approach of reducing capacity artificially (throttling).

This approach means that our congestion management techniques are precisely targeted and extremely effective – users in uncongested parts of the network are not impacted, and users in the congested regions are impacted as minimally and fairly as is technically possible.  The result is the preservation of quality of experience for the masses: maximum quality of experience for the maximum number of subscribers for the maximum amount of time

While it seems obvious that congestion management techniques should only apply where and when there is congestion, competitive approaches do not subscribe to this consideration.  “Capacity Control” techniques that are marketed as congestion management solutions simply drop packets.  For instance, a capacity control solution deployed near the top of the pyramid in Figure 3, might simply drop 20% of all packets of a certain type (e.g. limit bulk traffic to 20Mbps on a 100Mbps link).  On average, this approach delivers the benefit of reducing traffic on congested links by 20% at the cost of impacting every subscriber on the network by dropping 20% of the traffic on all the other links, whether or not they were congested in the first place.  This is approach is neither fair (since it impacts everyone) nor reasonable (since it impacts areas that aren’t congested). A link which had 150% of demand will not be uncongested in this model: it will now be @ 130%. Similarly, a link which was previously @ 80% of demand vs. capacity will now be @ 60%: a net loss of efficiency.

Compounding the confusion surrounding congestion management is the argument that another way to avoid congestion is through the use of usage management principles, like hard or soft monthly quotas (caps).  In this scenario, when a subscriber exceeds the monthly quota, the carrier might impose overage charges in the belief that the subscriber will then self-monitor and reduce data consumption.  The theory goes that if a carrier imposes monthly quotas on all subscribers, then the cumulative power of self-management will limit instances of congestion.  A related method is to lower the speed of users after some volume is achieved in a month. These principles are often disguised as ‘fair use policies’ (mimicking Sandvine’s Fairshare traffic management in name only).

Unfortunately, this approach is undone by the reality that instantaneous contributors to congestion are no more likely to be the network’s monthly heavier users (those who are exceeding their caps) than they are to be the other 99% of subscribers, so congestion remains.  Additionally, a monthly quota across the entire network is not narrowly-tailored to the links that are congested at the times they are congested. Why should a user be slowed down on a link which isn’t busy? The quota applies 24x7x365 on all links equally.

Usage management and traffic management each have their time and place. If your objective is to differentiate the value of service plans in order to maximise your revenue (a commercial objective), use usage management and volume limits. If your objective is a technical one, that of alleviating the effect of congestion on some links at some times in your network, use traffic management. Your users will thank you in both cases.