In May 2013, Capacity reported on the frantic pace of innovation in South Korea’s mobile telecoms markets. At the time, SK Telecom was readying its LTE-Advanced services for launch in September – an appointment that it kept earlier, in June.
SK Telecom and several other Asian operators are determined to drive 4G faster than anywhere else on the planet. KT in South Korea and Smart Communications in the Philippines followed swiftly after SK Telecom.
Europe then also began to catch up. In October, French operator SFR announced it had completed an LTE-A trial at its Marseille lab, getting download speeds of 174Mbps on the 800MHz and 2600MHz spectrum bands. In November, EE launched the first LTE-A pilot network in the UK. It is based around Tech City, the collection of startups (many of them working on telecoms applications), housed together in East London. It combined 20MHz of 1800MHz spectrum and 20MHz of 2600MHz, and promised a commercial launch in summer 2014.
By December, they were joined by Vodafone Spain, using spectrum in the 1800MHz and 2600MHz bands in its trials, and Telstra, which worked in collaboration with Ericsson to achieve speeds of 300Mbps on its 4G network by combining 20MHz of 1800MHz spectrum with 20MHz of 2600MHz.
Combining small cells and using carrier aggregation has, in these early stages, produced exciting results. The carriers investing heavily in 4G in Europe, the US and Asia have an evolutionary path that will take them through to 2020 at least – provided they can get the details right and convert exciting laboratory and small-test results into mass-market services.
Excitement grows
The trials, and SK’s LTE-A launch, have produced exciting headline numbers. Mansoor Hanif, the director of LTE at EE, points out that the benefits of LTE-A are already being felt around Tech City by all 4G users.
“By building a layer at 2600MHz separately to the layer at 1800MHz means we’re effectively doubling the capacity of the network. That has a benefit for everyone, not just LTE-A users,” he says. “We just got the first reports a couple of weeks ago. We’re excited because when we switched on at 2600, the load was perfectly balanced – everyone is getting an uplift. Even the iPhone 5, which can’t use the 2600MHz band, is getting better speeds.”
Currently used at 17 sites, including some outside Tech City, the throughput is faster than users – even those on the EE 4G service – are accustomed to, though it is well below the lab test speeds.
“About 25-30Mbps is average throughput. With 2600MHz, we’re seeing 40Mbps. But everything’s unknown. That’s what is so exciting. I told our engineers a year ago: get ready, because we’re ahead of our parent companies. Being a pioneer is hugely exciting and scary at the same time. There’s no-one to ask, because no-one knows. Being an engineer, there’s nothing better.”
While Hanif speaks excitedly of the potential for 4K broadcast content, or internet shopping via a mobile using holographic imaging (the sort of applications which developers at Tech City are attempting to build), he does not commit to a commercial launch in summer, no matter how well the trials proceed. That is because the fast download speeds are only a small part of the next-generation service, he explains. LTE-A is, at the moment, an exciting technology which has neither the handsets nor the applications to be commercially successful on a large scale.
“A handheld device doing 250Mbps at the moment would melt in your hands,” Hanif says. This, however, is not in the control of carriers.
A choice of compromises
The other sticking point for LTE-A – and one the carriers can do something about – is backhaul. Peter Jennings, CTO at MLL Telecom, works for a company that has been building backhaul networks for 22 years. He visited EE’s offices recently and found that he could download at 55Mbps on his iPhone. But he is not getting carried away.
“There’s no point in having a fantastic radio access, getting multi-hundred megabits, and having an inadequate backhaul,” he says. “The plans all the operators have in place will mean they could deliver LTE-A. We are trying to help them make their plans deliver in a timely and cost-effective way.”
There are, he explains, many trade-offs to be made. Better modulation brings higher speeds, but less resilience. Site sharing and network sharing has had an impact on how fast operators can introduce innovations, and how fast they can turn on the capacity they need. Investment ahead of a return is necessary, but it is in the interests of the purchasing department to keep that as low as possible. On the other hand, the type of traffic that LTE-A applications will use will have spikes and will be bursty, just like fixed-network IP traffic, so average demand is a poor guide to the sort of capacity a network will need.
Also, more operators are reversing the trend of the last few years and taking ownership of their infrastructure, so they have more control over service levels.
“We now offer design, support and installation, but as individual components,” Jennings says. “For LTE-A, the tower and the handset will have minor differences, but you’re looking at 10% difference at most between the equipment. But doing a good job of linking it back to your network changes the service. So for instance, downloading at 200Mbps to your handset but being restricted to 100Mbps backhaul would make a huge difference.”
The backhaul gap
In mid-2013, just as excitement developed around the early LTE-A trials, research sponsored by Tellabs and conducted by Strategy Analytics reminded us that the “backhaul gap” was not about to vanish. With operators spending, on average, 17.5% of their total operation costs on backhaul investments, Sue Rudd, director of service provider analysis at Strategy Analytics, suggested this was far too little.
“At today’s backhaul investment levels, operators could create a significant backhaul capacity shortage,” she says. Projecting forwards to 2017, Strategy Analytics research predicted that, for each $1 spend on backhaul above the 17.5% threshold, an operator could protect $4 in revenues, through less churn and increased demand. But the backhaul gap was widest in the three regions leading the way in LTE-A trials. In Asia-Pacific there would be underinvestment of $5.3 billion, which translates into a capacity shortage of 9.4 petabytes, while western Europe and the US were both $1 billion short (or 1.8 petabytes). No region was investing sufficiently in backhaul.
Now the good news: there is evidence that, in the all-IP backhaul environment that LTE-A dictates, capital investment in backhaul may be easier to fine-tune to protect service levels, according to Nan Chen, president of the MEF and co-founder of CENX. This is because all-IP networks allow the possibility of classes of service to protect applications like real-time video using Carrier Ethernet.
“The problem, even in today’s network, is that less than 50% of the transition to Carrier Ethernet has been done,” he says. “The OSS that supports TDM networks was not built to support the networks we will need. That has clearly been a challenge for the people who have ventured into 4G so far. It’s not just deploying the network, but managing it, provisioning faster, and providing accurate inventory for it. One of the challenges about Carrier Ethernet is the bursty nature of Ethernet, and it is hard to predict.”
Therefore carriers which were accustomed to upgrading capacity stepwise every couple of years may need to work more responsively. As demand spikes unpredictably in the IP environment, the most efficient LTE-A providers will use just-in-time provisioning, backed by OSS based on big data analysis, to predict and protect service levels.
Voiceover artists
One application in which this will be tested will be VoLTE, which many carriers see as one of the drivers of investment in LTE-A.
“Voice is very sensitive to delay and especially to jitter. You want to manage that very carefully. You certainly want to treat VoLTE differently to data. It will be a class that’s not mixed up with internet traffic,” Chen explains.
Steve Shaw, director of product marketing for mobile solutions at Juniper Networks, agrees that cells and handsets will not be the fundamental arbiters of LTE-A performance, but his prescription is even more fundamental: operators will be forced to completely re-engineer their core networks, compared to those providing 3G or first-generation LTE, to deliver applications.
“A lot of the RFPs we’re seeing now have specifically called out LTE-A. Our customers are starting to say, with LTE I see a customer getting 4, 5 or 6Mbps at the handset, I want to architect for 100 or 200Mbps at the handset,” Shaw says.
“For example, security in LTE is handled in a hub-and-spoke type environment. Encryption runs from the cell tower to the core. But in an LTE-A environment, you need to push your security much closer to the cell sites. You can’t take the traffic all the way back, hairpin it around and then bring it back again,” he adds.
That means using software-defined networking and smaller, software-configurable data centres to push security much closer to the cell tower itself. It also means replacing proprietary network equipment. Shaw sees two sets of responses.
“Many realise the need, but securing the access network has not been a high priority,” he suggests. “We’re almost supplying a reason to them why it’s too much work to do. But the ones who have implemented security in LTE, for example, realise what they need to do. The ones that haven’t are almost using this as an excuse to keep not investing.”
Closer to the edge
When customers are getting 200Mbps to their handsets, both backhaul and the core have to be reimagined, says Shaw.
“How do I run packet core elements closer to the edge? How do I create data centres that are much more distributed. We were talking to one service provider who said, ‘I have five large data centres in my country, but if I was going to architect my network to support 200Mbps, I’d probably have to push my network edge so far out that I would have 200 small data centres positioned throughout my network’.”
While operators have a regulatory burden that prevents them from operating with the agility of an Amazon or a Google, Shaw is “always optimistic” about their response.
“Operators want to push virtualised services closer to the edge, when that makes sense. Security, load balancing, video optimisation: run them as virtual appliances. There are profound impacts across the network. Getting to this point where there is a more cloud-like approach, where services can be put in as virtual machines in two weeks instead of six months, we’re going to see a lot more innovation. There’s pent-up innovation in these service providers,” he says.
“We’re building the mobile internet,” Shaw adds. “LTE was version 1.0, and LTE-A is version 2.0.”
