Home
Market research, financial analysis and consulting on wireless technologies and services
 
From FierceWireless

In a recent report, I argued that data caps are not effective as a traffic management tool as they may encourage peak traffic, and thus increase congestion, instead of containing it. Fairly enough, one question that I often get is what works then, if data caps don't. Mobile operators have different tools--Wi-Fi and small-cell offload, policy, content optimization, QoS-based tiering to name a few--and they will have to employ many of them to manage traffic effectively.

One can argue that since the time-of-day usage patterns are known (Figure 1) and are relatively constant across markets and time, most of the benefits of traffic management can be extracted by simply activating tools during the evening peak hours. Yet, if we go beyond the average and look at different location types (Figure 2) we see that the traffic distribution through the day varies greatly. As a result traffic management tools that work across the entire network are bound to have only limited efficacy. They either have to be too broadly used to capture peak usage all different location types, or too narrowly used and unable to effectively manage traffic at peak time in a subset of locations.


Figure 1. Time-of-day network mobile downstream traffic in North America. Source: Sandvine


Figure 2. Time-of-day traffic distribution by type of location. Source: Nokia

Eventually, to fully benefit from traffic management tools, operators will have to move to real-time, cell-level traffic management in the radio access network (RAN).

And moving one step further, operators need to actually act ahead of time, using predictive data to prevent congestion, especially when due to unexpected traffic spikes. Acting after congestion arises is still beneficial, but the true goal of real-time traffic management is to prevent congestion from happening in the first place, by observing traffic patterns and enabling traffic management tools as a cell approaches full utilization.

Most mobile operators I talked to would rather avoid to do this and reasonably so. Tracking and acting traffic in real-time at the cell level inevitably adds complexity and today traffic data is available but only offline.

As traffic patterns are mostly consistent through time, offline traffic information is useful to predict congestion times and plan for capacity upgrades. However, it has only limited value to predict and manage traffic peaks--especially unexpected ones, e.g. ones caused by a game or other public event--to avoid congestion.

Today operators are just starting to implement traffic management, policy and content optimization to increase the efficiency of their networks. These tools allow them to use their network resources more efficiently, i.e. to pack more data within their existing pipes. The cost per bit goes down, and, if properly implemented, QoE is improved by sharing resources more fairly among subscribers.

However, this necessarily comes at a price. For instance, video optimization may improve the subscriber experience at peak hour, when reducing the amount of traffic downloaded may prevent video streaming from stalling or skipping frames. During off-peak times, when there is available network capacity, video optimization does not have any beneficial impact on the subscriber experience, but it adds overhead to the network. Operators may decide to only use video optimization in locations and at times in which traffic is high, but this approach lacks the flexibility to deal with sudden or unexpected traffic changes. Using real-time traffic information, the operator can limit the use of video optimization (and the associated processing costs) to the times when it is needed.

Similarly, enforcing policy when needed maximizes its beneficial impact on network resources--typically by reducing peak traffic--but minimizes its restrictive perception on subscribers. For instance, there is no need to throttle a user connected to a cell that is not at full utilization. In fact, it is preferable to encourage the subscriber to use its data allowance in off-peak times and locations, as this may increase the overall network utilization, if we assume that as a result this subscriber will reduce its use of the network at peak hours, allowing other subscribers to benefit from limited peak capacity. If the subscriber simply increases its overall usage levels and does not reduce its consumption during peak times, the operator will still be able to count on a happier subscriber, but it does not face any additional cost as off-peak, otherwise unused capacity is effectively free to the operator.

At which point the granularity of real-time information becomes excessive, eventually creating so much complexity and overhead that the benefits decrease or disappear? This is a question that is still open, as traffic management is still in its early stages of development--and in its even earlier stage of adoption. Monitoring the costs/benefits tradeoffs, both from a financial and performance perspective, will be necessary to ensure that operators achieve the desired balance.

Furthermore, it is imperative that the resulting traffic management strategy is sufficiently simple and transparent that it can be easily communicated to subscribers, and that they perceive it as a way to make the distribution of network resources fair--rather than the result of the capriciousness or greediness of the mobile operator. Again, this is another area where there are no established answers from experience--and in fact different mobile operators may choose to go different ways. And here lies one of opportunities that real-time traffic management brings to operators. They can choose how to go about optimizing their networks and find different solutions. Even as they use the same 3G or 4G technology and have comparable speed rates, they can differentiate from each other in the way they use their network to meet the expectations of their subscribers.
 
 
Bookmark and Share
 
 

| | | | |