5G: wireline as much as wireless, and not all about constant connectivity – Rethink Technology Research

Rethink Technology Research
By Caroline Gabriel
Published It is more meaningful, in many scenarios, to think of 5G as a new generation of connectivity, or a broader architecture, than just a radio upgrade. Fully virtualized, even sliced, networks; multiple air interfaces; dynamic access to multiple spectrum bands – all these and many more have their roots in the 4G HetNet, but will come to fruition in the next generation. However, that thinking can be extended further, to make the case that 5G will be as much about wireline as wireless, and may not always be much about connectivity.
Wireline advances, such as high quality fiber and eCPRI, will be essential to provide backhaul and fronthaul for the dense, virtualized 5G RANs, but the ‘5G is wireline’ argument goes further than that. Operators are increasingly looking to a converged, seamless and flexible architecture in which fast wireless and wireline links are selected dynamically according to the requirements of a particular user or service (or network slice). Mobile connections will onload to fixed ones, which could be wired or wireless.
And even in the case of mobile-enabled devices, many 5G use cases will require less mobility, and indeed less connectivity, than 4G. Increased data rates and lower latencies – cornerstones of the 5G vision – will tend to push network intelligence and processing to the edge, along with the distribution of cloud computer and storage resources. Internet of Things applications with fast response requirements such as vehicular safety, as well as high speed video, will be better served from an access point and a processor close to the user to minimize traffic back to the macro network, the mobile core and the cloud.
Evolved packet cores (EPCs) and content delivery networks (CDNs) will increasingly be distributed and virtualized and many connections will be made over very localized networks – not just cellular small cells, but low power Bluetooth or Zigbee clusters linked by a gateway to the 5G link (where they are attached to the wide area at all).
In some applications, then, there will be a sharp swing away from the assumption that 5G networks will become increasingly controlled from a centralized cloud platform. Instead, they will be made up of a huge collection of localized, sometimes application-specific sub-nets which spend most of their time moving data around in their own edge-based location. For some services, we could foresee a return to, in effect, the old days of batch processing, with data generated from IoT or video clusters being returned to the central data center just a few times a day.
These are the usual swings of the networks market – from centralized to distributed to centralized again, as seen in corporate networks (mainframe to client/server to cloud) and in broadband topologies. But in the 5G era, they will have a profound influence on the cost and deployment norms of the cellular and converged networks.
Operators will be looking at far more multi-layered networks in the 5G era and will have to make key decisions about how they combine wireline and wireless for the best outcome in terms of customer experience and cost of delivery. Old assumptions will break down. In fixed wireless, fiber will always beat cellular for sheer speed, but the former will be fast enough to appeal to users who also want a single connection which is mobile too.  On the other hand, advanced in fiber and in copper (G.fast) will allow wireline to penetrate in areas which were not cost-effective before, such as multi-dwelling units in low-ARPU locations, or rural regions.
The most important aspect of ‘5G’ will be the development of virtualized, dynamic platforms which free operators from making either/or choices about different technologies and allow them to tap into different combinations of connections for different scenarios. These may be their own multiple networks managed from the cloud; or they may rely on building just one type of network (even WiFi) and then using a third party cloud service to supplement that by buying capacity on other wireline and wireless networks as required, paying on a per-usage and marketplace basis.
Such an approach would transform the economics of telco and cloud service provision, and though many pieces need to fall into place before the flexibly converged network is a reality, early signs are seen in the tests and trials of pioneers like Korea Telecom or Telefonica, while some of the technologies which are emerging now will become enablers.
For instance, SDN (software-defined network) controllers coupled with network slicing platforms and marketplace charging/roaming mechanisms sound futuristic, but are already a reality in certain applications. AT&T’s on-demand services for enterprises are wireline but will encompass mobile in future and point the way to slicing; BandwidthX has pioneered a marketplace system for trading WiFi capacity but has recently extended its reach to cellular; companies like Quortus, as well as majors like Nokia, have virtualized, localized and IoT-focused EPCs in commercial use; Huawei has shown flexible spectrum usage with its innovative CloudAir.
Many other innovations are in the works, promising to fill gaps in the converged, virtualized networks picture over the coming years.
And if 5G is to be wireline as well as wireless, it is important to look at the developments in that industry, rather than staying in a cellular niche. Last week, for instance, the cable industry gathered in Cologne for the ANGA COM summit, and showed a variety of technologies which could vye with fixed wireless – currently such a fashionable predicted use case for cellular 5G – in the affordable last mile.
Indeed, with MoCa (Multimedia over Coax) moving out from in-home environments to the last mile, and DOCSIS 3.1 and G.fast providing alternatives to pure fiber – which can build on existing cable and copper systems – it could be argued that fixed wireless’s bright new star will be extinguished as quickly as it was in the LMDS and WiMAX days. If wireline technologies can outpace wireless in cost and flexibility; can integrate with a mobile or portable experience via a shared operator core or a WiFi-first approach; and can come close to pure fiber in speeds, where is the space for fixed wireless, except in ultra-rural areas, or for operators like Verizon and AT&T with a very specific need (in their case, to move beyond their wireline territories)?
Fixed wireless may remain a niche application of 5G, but that is not the real point. The point is that operators will need to build 5G networks which combine fiber (or other wireline alternatives), with cellular for mobility, and with localized in-home or edge-based technologies, some of them expanding their reach into the last mile (like MoCa and WiGig). The more flexibly they can do that, in a virtualized environment, the more they will be able to balance cost with user experience to achieve competitive economics, and to be well-positioned for unforeseen new traffic patterns or use cases in future.
Some wireline advances which may squeeze wireless 5G:
MoCa (Multimedia over Coax) is trying to move from being an in-home LAN connection, to being used for last mile broadband access, as demonstrated at the ANGA COM event in Cologne last week.
The last mile has been shortening as the twisted pair of the telco has had to inch ever closer to the home – what was once 6.5 kilometers in ADSL is now shrinking to anywhere from 1,500 feet down to 70 feet in G.fast. MoCA Access, which launched at ANGA, after being mooted a year ago, is only for coaxial cable and reaches 150 meters, which makes it fine for most US homes, but a far bigger bet may be the transport for multi-dwelling units (MDUs) in Europe and Asia-Pacific.
This MDU market has become one of the bones of contention in broadband over the past 12 months. As long ago as February 2014 a trial with Korea Telecom let G.hn into this same MDU coaxial broadband domain, although this technology, we understand, lost out to a cheap Ethernet system last year. AT&T has talked about a system which uses G.fast chips with coaxial cable, and Sckipio has announced such a system, operating at around 1Gbps, and is promising a 212 MHz version at almost double that speed. The vendor is  pushing out a different system with Calix, over twisted pair. There is even a version of this G.fast system which coexists with millimeter wave wireless backhaul from up to 2.5 miles away, which Sckipio has perfected with Siklu for MDU delivery.
The MDU is the last great challenge for broadband, and many things have been tried, among them Ethernet over Coax and bendable fiber to help with retrofitted installations. But the most frequently discussed systems will use the wiring that’s already installed. Most MDUs have a twisted pair, installed to bring in telephony, and a coaxial cable originally deployed to bring in analog cable. Both tend to reach with fiber from the basement up, although there are instances of dragging backhaul in halfway up a building from overhead cabling.
Most MDUs have already been fitted with two-way repeaters, so that they can handle broadband, but that too is not a given, certainly not in the poorer parts of the world. So there is now a race against the clock to landgrab over 500m or so homes which do not even have broadband available to them, over the next 5-10 years, and the system that wins, will be able to use this base as an economic platform to attack other types of homes.
This could be a target for fixed wireless systems (one of Ruckus’s early differentiators was a WiFi system targeted specifically at high apartment blocks, mainly in Asia). But the wireline alternatives are proliferating and getting cheaper and more flexible too.
MoCA Access is based on MoCA 2.5, a new release which uses five 100 Hz slices each way in spectrum from 400 MHz to 1675 MHz, multiplexed to deliver around 2.5Gbps in the downlink and 2Gbps up, and devices are available as of this month. The way MoCA sees this working is for the fiber to be split typically four ways, for an MDU of about 24 homes. That puts six homes on each path, which means they share 2.5Gbps down and 2Gbps up, and can be backhauled on GPON or any other fiber. One of these access systems can take signal up to 150 meters away from the backhaul split without a repeater, and can drop it into up to 63 devices on a point-to-multipoint, two-way topology. So those six homes can have around 10 modems each to connect directly into something like a TV, or can be split further using WiFi. The IEEE 1905.1 protocol supports that WiFi connection and there are plenty of devices already available.
It also supports eight classes of traffic for different QoS levels, so traffic can be layered.
MoCA suggests that latency is less than 5ms which also makes it perfect for 5G backhaul – the most immediately obvious area where cellular and wireline communities will meet, and where various wireline technologies may see a new opportunity.
“MoCA Access leverages our core strengths in high performance, reliability and no new wires,” said Charles Cerino, president of the MoCa Alliance. “Service providers around the world can take advantage of a proven high speed technology that is designed for all future installations. It is also the perfect complement to a wired backhaul architecture for upcoming cellular technologies such as 5G as it has very low latency.”
It is amazing how resilient MoCA has been, given that it began life with one chip provider Entropic, and it was years before any others joined the game. Broadcom was next to come out with a chip, but it had yielded several years of headstart to Entropic.
Interestingly Entropic was acquired by MaxLinear, which went on to acquire Marvell’s G.hn technology as well, as well as specialist mixed signal business Exar, which works closely with mostly large Tier 1 Chinese customers. This may point to potential to merge the two systems. G.hn has some acceptance in China, so buying Exar and Marvell’s G.hn will make MaxLinear a trusted supplier there. It can then supply whichever component the Chinese like the best for any given application – MoCA Access for access, and perhaps the powerline version of G.hn for in-home backhaul back-up? If a Chinese partner prefers DOCSIS, MaxLinear is also suited for this because it provides RF components to create multiple DOCSIS tuners in one piece of silicon. It has all bases covered.
While US cable equipment suppliers may remain rigidly faithful to DOCSIS, this is not necessarily true of Chinese cable operators, and they could switch to MoCA Access, if it can promise cost and performance advantages over DOCSIS.
Of course, the supporters of DOCSIS will not surrender easily. Also at ANGA COM, a line-up of major cableco suppliers were announcing updates to their DOCSIS 3.1 product lines. Arris (soon to acquire Ruckus), Cisco and Casa Systems all unveiled new offerings to accommodate the rising operator demand for DOCSIS 3.1 products.
A notable difference this year centered around new types of virtualized CCAP (Converged Cable Access Platform) technologies. CCAP is the combination of the network server which drives QAM TV with the CMTS (Cable Modem Termination System), which in turn drives DOCSIS connections. Almost all DOCSIS 3.1 is being delivered in CCAP format, and virtualized CCAP has been talked about for the past three years as a method for one of the fastest replacement technologies ever for cable – with a smaller hardware component meaning little economic reason for delay.
Cisco announced its new Infinite Broadband Remote PHY service, combining the two CableLabs standards, Remote PHY and DOCSIS 3.1, and claiming symmetrical speeds of 600Mbps. Cisco is hyping up Remote PHY as the technology that forms the foundation of its strategy towards virtualization and full-duplex DOCSIS. It builds on its existing cBR-8 converged broadband router and GS7000 node platforms, expanding cable capacity, reducing power consumption, and providing site size savings. Cisco has a large scale customer deployment with Altice, which has rolled out Cisco’s cBR-8 with SFR in France, and last year began deploying the technology in the US.
Cisco’s product marketing director, Daniel Etman, wrote in a blog post, “To compete with pure fiber, cable operators must scale bandwidth to meet the needs of the market. In the past, they relied on two ‘tried-and-true’ techniques to achieve this end-segmentation and spectrum management. Unfortunately, network segmentation to improve bandwidth distribution has all but reached a physical limitation with legacy CMTS platforms. And adding more ‘boxes’ in already cramped hub sites is not the answer. These kinds of network changes are complex, expensive and don’t solve the long-term problem.”