This niche event (approx 50 delegates, majority from European operators) has quite a different feel to some of the larger conferences that attract bigger marketing budgets and carefully crafted messaging. I find you can sometimes get good insights into what's really going on "at the coal face", and what planning/optimisation engineers are really thinking. It seems this topic is gaining importance while it wrestles with more complex network architectures, more demanding customers and increasing cost pressures.
ThinkSmallCell chaired Day 2 of the event but we've summarised the key themes and highlights from both days.
QoE vs QoS
The opening presentation from VIPNet's Service Quality Assurance manager was excellent, setting out the scene on factors affecting management of end-user quality. VIPnet is a Croatian fixed/mobile network, part of the larger Telecom Austria group, and accommodates many tourist visitors during the season. Hrvoje Jerkovie clearly laid out the difference between QoE (Quality of Experience) and QoS (Quality of Service):
- QoE is what the end user feels about the service quality. It may be affected/impacted by a strong marketing campaign about how good the network is, not only by better engineering. There is often a lag after making network improvements before these are recognised by customers. Those leaving after a bad experience take time to be convinced that performance has improved.
- QoS is what the network performance engineers can measure and report on. KPIs (Key Performance Indicators) are reported separately for network oriented statistics (such as call drop rates, handover success rates, average data rates etc.) and service oriented statistics (such as web page download time, voice call setup time, YouTube video success rate. Delegates generally focus on a few (say 10) parameters, rather than 100's.
Data isn't a single service anymore
With this QoE in mind, it was said that users don't really think about data rates or specific KPIs, but instead the performance/responsiveness of each App they use. For YouTube it might be how quickly the video starts showing (assuming it then doesn't stutter). For a webpage it might be how quickly it renders. For a multi-user game, it might be latency with other users also online.
It seems the top issue from customers now isn't (lack of) coverage anymore. We usually have some bars showing almost everywhere we go. Instead it's slow data rates, or lack of responsiveness to our Apps, whether downloading email or looking up the weather forecast. This has become more specific for each of the different Apps that we use.
Payload versus Signalling
A few years ago, there was a lot of concern across the industry that excessive signalling (the chatty message exchanges between Apps and their Cloud servers) might overwhelm networks even more than the volume of data. This was relieved by features such as Fast Dormancy, which reduces the number of messages significantly, but it continues to consume substantial resources. While usage is charged by the Gigabyte, heavy signalling traffic isn't really charged at all. Turkcell explained the issue in depth and I got the sense that this is mostly under control, but that some handsets/Apps (even large well known brands) still wake up their network connection every minute.
When this goes wrong, as it did with a WhatsApp server outage some time ago, it can cause signaling storms that seriously impact the network for all users.
VoLTE isn't for all
There were contrasting views about Voice over LTE. The VIPnet speaker simply didn't see the attraction, quoting (1) few supporting devices as yet, (2) not yet completely confident in the technology, and (3) significant disruption/duplication for back office systems. They already have HD Voice on 3G which they thought is at least as good, and possibly even better than VoLTE. So despite having been one of the early IMS customers, they look likely to use CSFB (Circuit switch fallback, carrying all voice calls on 2G or 3G) for some time. Other European operators told me they were in advanced trials and plan to launch fairly soon.
Tektronix ran through all the implications of VoLTE, including how to measure and guarantee performance. Their view was that you do need good blanket coverage of an area (e.g. a city) before launch. In response to my question, it was confirmed that a home network operator with VoLTE could offer that service to their customers when they roam abroad, even when the foreign 4G network didn't. So you may well be able to enjoy VoLTE in Croatia as a tourist in future.
Some countries won't be adopting 4G quite as quickly as others. The Belgian speaker reminded us that they don't have handset subsidies in that country, with 65% of devices are still 2G only. One Asian delegate mentioned his 93% GSM user base.
Not all yet singing off the same hymn sheet
Consultancy firm Omnitele works with both technical and marketing teams in operators. Mikko Valtonen said there is often a lack of alignment, with CTO and CMO teams not using the same common language, KPIs or even business objectives. The CMO may be worried about poor perception by their customers (QoE) while the CTO may be reporting green lights on all their network quality metrics (QoS).
This can be partly attributed to the use of many different, disconnected systems for performance reporting. Infovista championed the integration and use of a single common reporting system.
A structured and holistic approach to prioritizing future investment
Ashley Benyon from KPN Netherlands described how they use a much deeper analysis around prioritizing and justifying each new network investment. They consider all the options (expanding with more basestation capacity may not be the only choice) and validate the RoI (Return on Investment) for each project. They want to incorporate customer centric information and achieve financial transparency when prioritizing capacity investments.
The case study given was the main RAI railway station in Amsterdam, which is covered by a single basestation – their busiest in the country. They can choose from lots of spectrum: 5x 3G carriers and LTE in the 800, 1800 and 2600 bands.
The station is served from a single three sector macrocell site using 3x 3G carriers and 1x 4G 800MHz carrier, but still can't meet forecast traffic loads.
Analysis showed that much of the traffic comes from 4G capable iPhone 5 (not 5c or 5s) that use 3G because that phone only supports 4G at 1800MHz. So they could either deploy an 1800MHz LTE carrier to cater for that traffic, or subsidise more rapid takeup of the 5c/5s smartphones. I did ask if Small Cells or other technologies were an option but KPN don't yet use them widely outdoors yet and prefer to stick with a standard set of equipment configurations. In future, I'd guess that they would where the business case proved it economic.
Avoiding high profile problems
It seems to me that a few high profile locations and events are still accorded far too much attention. Whether it's the Superbowl in the US, the World Cup in Brazil or EXP2015 in Italy, operators can't afford the bad publicity that comes with outages or poor performance. Large sums are spent on specific venues and it seems to me that several engineers were still looking for the highest performance solutions in these extreme situations. This also includes handling the enormous volumes of signaling and handovers as vast numbers of people move around.
Consumption continues to grow like a rising tide
There was a gentle and healthy rivalry between the EE and 3 UK delegates. Last year's traffic figures (to June 2013) showed that in the UK, 3 carried 43% of all UK mobile data traffic with EE on about 27%. Since then, EE have launched 4G so would expect to have leapt ahead. Three couldn't reveal their traffic growth figures publicly, but did say that it continues to grow relentlessly every month. Whether EE/3 together will retain 70% of all UK mobile data traffic when this year's figures are published remains to be seen.
Orange reported a substantial change in traffic source, which is now 70% from smartphones and only 6% from USB sticks/computers. They see very little from tablets, which appear to be mostly used directly on Wi-Fi. Some users turn off Wi-Fi when LTE is available (even at home), and the much faster uplink speeds of 4G (up to 20x faster, and sometimes even faster than download rates) are affecting the uplink/downlink ratio.
I hadn't heard of Hutber's Law before. Mark Gilmour from 3 UK revealed this somewhat cynical view where "Progress doesn't necessarily mean improvement". Several operator delegates remain sceptical about many of the latest industry initiatives, such as SDN, NFV, VoLTE and even SON (Self Organising Networks). This view comes from senior engineering managers with decades of experience who still need convincing. However, when I asked if Small Cells would also be included in this category, I was firmly told no - Small Cells are one of the few absolutely essential tools needed in the future to deliver the quality and capacity users demand.
Phew, that's a relief then.