Speed and efficiency of the radio interface are the primary characteristics of 4G/Long Term Evolution (LTE) – and rightly so, as that's how mobile technology generations are generally defined – but equally significant is the shift to an end-to-end IP network architecture. It's not for nothing that the formal name for LTE has, for some years now, been Evolved Packet System (EPS).

The implications of all-IP mobile networks are far-reaching. The ability to support a diverse mix of high-value, real-time, and lower-value best-effort services over a common IP infrastructure is vital to those mobile operators seeking to achieve an attractive blend of revenue and profits.

Yet this transition takes the industry into largely uncharted waters. Elimination of the 2G/3G circuit-switched domain used to support voice and SMS (which constitute 80 percent of operator revenues today) throws up one obvious challenge: With no circuit-switched domain to fall back on, mobile operators are being forced to think about how to implement quality of service (QoS). And once we accept the need for QoS – and by association, policy control – thoughts turn to how to use it to open up new opportunities.

(For an introduction to EPS QoS, I would recommend the book SAE and the Evolved Packet Core by Magnus Olsson and colleagues.)

The starting point for QoS in LTE is the EPS bearer model defined by the 3rd Generation Partnership Project (3GPP) . This is the formal QoS scheme in which the network can allocate service data flows (SDFs) to specific "EPS bearers" with quality class indicators (QCIs) assigned according to available network resources and decisions made by the policy charging and control (PCC) architecture. It's a vast oversimplification, but the ability to treat different applications and users to different service levels is inherent to the LTE specifications. In theory, there's a lot that could be done, both commercially and creatively, with this capability.

However, two points not covered in the standards or textbooks merit further attention: Does the EPS QoS model work in practice? And is there a commercial reason to build this complexity into the network?

Certainly there are reasons to be cautious. Advanced functionality specified for the primary intent of monetizing user applications and services is not inherently a bad thing – and one could argue that without it, the industry has nowhere compelling to go. But what looks right on paper may take much longer, and end up being less convincing, in reality. Echoes of IP Multimedia Subsystem (IMS) ring loud.

Does the technology work? At this stage, no one can really say with 100 percent confidence. While the raw throughput and latency metrics of LTE are meeting or exceeding expectations, experience of rapidly setting up, tearing down, and assigning QoS to sessions in loaded networks is still limited. Even the leading vendors cannot yet show data that proves, unequivocally, that the EPS bearer model works under stress.

Feedback from operator testing is also inconclusive. Mindful of their experience with 3G, operators are working to simplify LTE technology as much as possible – with the side-effect that, when they have begun to stress-test performance of delay-sensitive applications and QoS mechanisms, those operators have discovered that it is far from straightforward, and may even perform worse than in mature, stable 3G networks. The message from operators is that there's work to be done, to put it kindly.

Then there's the commercial model. Voice is an obvious application that can be monetized and requires stringent performance in the network. But elsewhere, questions abound: Is there an equivalent of a triple-play IPTV service in the mobile domain? Do emerging machine-to-machine applications need guaranteed service? Could an operator really sell QoS to an upstream video provider? Would industrial or corporate customers be prepared to pay for a service-level agreement (SLA), even if the operator could deliver one?

If these questions can be satisfactorily resolved, QoS can potentially enable operators to create a robust service mix that will maximize yield from their LTE network investment. This is a key theme of the Light Reading Webinar I recently hosted on "Capturing New Opportunities in the Transformation to All-IP Mobile Networks" and is of direct relevance to my upcoming report on voice over LTE, as well as a raft of other Heavy Reading research into next-generation mobile network technology.

— Gabriel Brown, Senior Analyst, Heavy Reading

shygye75 12/5/2012 | 4:37:20 PM
re: LTE & QoS Gabe -- Given the QoS levels of today's mobile networks, is it reasonable to expect operators and suppliers to work out QoS benchmarks for LTE at this stage?
Gabriel Brown 12/5/2012 | 4:37:16 PM
re: LTE & QoS

Good question. QoS in today's 3G networks is more about, either you get coverage/service or not. Sometimes it's fast, sometimes it's slow.

From that viewpoint, setting benchmarks for QoS in LTE networks is not an immediate priority. But it will become important. LTE voice will probably be the trigger.

Progressive operators are already some way down the evaluation path.

Perhaps the best example, and one that's semi-public, is T-Mobile, which showed results of an LTE voice trial (using VoLGA) at Mobile World Congress. The results in terms of call set-ups and so forth were very impressive.


shygye75 12/5/2012 | 4:37:15 PM
re: LTE & QoS Thanks, Gabe. How will the benchmarks for LTE QoS be set, and who will be setting them?
Gabriel Brown 12/5/2012 | 4:37:11 PM
re: LTE & QoS

The actual QoS classes, and associated performance metrics, are defined by 3GPP.

Benchmarking the performance of equipment, and how well it performs under stress is down to operators and vendors for the most part.

Third party testing - e.g. by the Multi-Service Forum - is generally focused on interoperability rather than on the performance of the equipment.


shygye75 12/5/2012 | 4:37:10 PM
re: LTE & QoS Interesting that "quality of service" initiatives don't appear to have any input from end users. Does that mean QoS isn't the right label for these programs?
Sign In