Operator Perspective: Assuring QoE for Next-Generation Services

Posted by Mae Kowalke on Wednesday, March 02, 2016 with No comments

How are mobile network operators viewing and addressing quality of experience (QoE) for next generation services like VoLTE? A recent RCRWireless report took on this topic by including insights from interviews with several operatorspresented in a larger context with vendor observations as well. A summary of the main operator-focused topics is summarized below. 

What is 'reliability'?

Reliability means customers can do what they want, when they want, on their mobile devices. As an example, for a voice call there are three parts or steps to reliability: 1) connect the call, 2) keep the call connected, and 3) assure expected speech clarity during the call. All three are needed to deliver a service that is experienced as being reliable. 

With the tools and practices currently available, it is hard for operators to focus on reliability (and QoE) at a micro level. Therefore, they are starting with fairly broad approaches, such as:

  1. Consider the the tolerance limit of each service and focus first on optimizing those that have less tolerance. (Voice has less tolerance and therefore higher priority than browsing or email, for example.)
  2. Optimize the network with the aim of delivering an amazing experience for the vast majority of customers at each location.
  3. Determine which services the majority of customers are using, and how often. Start optimization efforts there. 

How does 'customer experience' correlate with 'network performance?

Sometimes, the network seems to be performing well (all elements are up and working as expected) but customers still have a bad user experiencefor example, with in-building coverage where customers segments can and do have varying degrees of tolerance regarding coverage quality and reliability. 

Instead of looking at average customer experience, operators need to examine how experience is distributed across different segments, and how that perception connects with profitability. 

Further, network engineers traditionally have relied on KPIS, which are historical measurements and usually are averaged. But that model doesn't work when a subscriber's experience is taken into account, for specific applications, using real-time data. It's necessary to correlate subscriber data and KPIs, but they often do not map easily. 

What is 'real-time?'

Traditional network monitoring uses historical data. But a subscriber's QoE from a week ago isn't helpful when the goal is to optimize the network in 'real-time.'

The definition of 'real-time' depends on the context. For the overall network, a 24-hour window is good enoughfor now. Do that first, and then starting lowering the time resolution. During special events, 10-15 minutes is necessary. 

During special events, it's possible to have necessary capacity but if it's not optimized, there will be QoE issues and an unbalanced network. 'Real-time' data and insight is needed to know when to trigger parameters to guarantee the maximum possible QoE during different traffic level periods.  

How should network performance and QoE be measured?

We've left the voice era and are in now in the data era. This means operators must stop managing traffic using a siloed approach, and instead need end-to-end network visibility and planning. 

The trick is, each part of the network has its own type of KPIs and measurement systems. Measuring access and core performance is pretty well-established and doable. Measuring QoE is a whole different ballgame. Operators are trying out different types of probes and DPIs to get proper QoE measurements, but as of yet there's no established way to do so in an end-to-end manner. 

Making this more challenging is the fact that more and more traffic is becoming encrypted, so it isn't possible to see what type it is. Without being able to differentiate different types of traffic, how can operators measure the associated QoE? 

One method is to look at the traffic volume, and use an algorithm to deduce that it's, for example, streaming video. From that, an operator can guarantee a specific throughput. 

But what's really needed is a way to take raw data from the network and use that to determine quality of experience. Since users won't tolerate on-device apps to report on QoE (which would be ideal from the operator perspective), other options must be considered, such as field testing and probes. 

Why are QoE and QoS so important?

This might be obvious, but is worth saying anyway: from a business perspective, quality of service is a way of monetizing data. Good QoS is a must to offer the level of QoE that keeps customers on a specific operator's network. This means end-to-end QoS. 

Ultimately, achieving necessary QoS and QoE requires simplifying the access network; current architecture isn't sustainablefrom landlord point of view, power point of view, or capacity point of view. 


Categories: , , ,