What is an Open RAN?
The concept of open goes beyond just standards interoperability. A RAN built by Nokia, Ericsson or Huawei will be interoperable with any device, any core, any transmission network attaching to it – due to its conformance with standards. That is a standards-based network. But the software and interfaces remain either proprietary or optimised by the individual vendor, and are often tied to the underlying hardware layer. Some might argue this is still the case even in a vRAN deployment that, whilst it lifts software away from underlying hardware, doesn’t open the Remote Radio Head (RRH) to vBBU (virtual BaseBand Unit) interface.
The vision of the Open RAN is that it is open within the RAN, with the interfaces and operating software separating the RAN control plane from the user plane, building a modular base station software stack that operates on common-off-the-shelf (COTS) hardware, with open north- and south-bound interface. This architecture would mean that baseband units, radio units and remote radio heads can, in theory, be assembled from any vendor and plugged together to form a network.
FROM SECRET SAUCE TO OPEN SOURCE
Why would operators be interested in this? Well, first it gives them that multi vendor flexibility within the RAN itself – the ability to adopt best of breed in a network, and reduce reliance on a single vendor.
More than that, it devolves the networks away from the reliance of vendor specific “secret sauce” – say in CPRI implementation, in hardware acceleration, in processor and in chipset optimisation.
The cost implications of both these advantages are huge. The first offers a negotiating position – a viable “other” to the current vendor pricing model. The second builds on that by reducing dedicated hardware cost. Open source designs for the radio software could minimise costs still further.
However, the technical consequences are also sizeable. Radio network processing is intensive, real time, and complex. It has relied on the optimised software and hardware capabilities – working in tandem – of specialist vendors. It has seemed very unlikely operators will put their hard won network reliability onto white box IT servers running software from new entrants.
Open RAN, therefore, has seemed like a technical project to date, driven by niche vendors partaking in operator research projects such as XRan, and by the Facebook-backed TIP as a means of discovering low cost ways to connect the unconnected.
It has been unclear whether operators have been truly committed to commercialising Open RAN technology, or if they liked to use the threat of it to keep their incumbent vendors honest.
Yet at MWC there was not just a reaffirmed operator commitment to Open RAN, but a clutch of vendor moves to support that deployment model. Is it happening for real? Are operators really behind it?
The first big operator news came with the announcement that the xRAN Forum and C-RAN Alliance would be merging into something known as the Open Ran Alliance. Operators including AT&T, China Mobile, Deutsche Telekom, NTT DOCOMO, and Orange. ORAN Alliance (formerly xRAN Forum) is looking for “virtualised network elements with open, standardised interfaces”, and added that “open source and open whitebox network elements will be important software and hardware components of these reference designs”.
The Telco Infrastructure Project announced extensions to its Open Cellular programme and the addition of an Open CrowdCell project. CrowdCell is a concept based on LTE UE relay architecture that Vodafone has to this point developed with Huawei. The TIP project sees it move the concept to generic hardware and open source software designs. Companies such as Amarisoft (also backed by Orange funding in fact) and Quortus are involved.
The Open RAN project dis-aggregates current RAN platforms by separating hardware and software, to implement on COTS platforms with open interfaces.
Vodafone’s demos mentioned the involvement of Baicells (with a product at the booth), Parallel Wireless and Altiostar, as well as Radisys and Lime Microsystems. Lime has previously been involved with BT/EE on its trials of programmable radio using open source software. Andy Dunkin of Vodafone said that the company was looking for commercial product this year – with the rural coverage use case being the first. That would then feed revenues back to vendors to be able to expand their efforts, he said.
Bruno Jacobfeuerborn, until recently Deutsche Telekom’s CTO, said at a “Funding Network Investment” session that operators need to shift from buying boxes to buying programmable eNodeB’s on standardised hardware.
Really, on the RAN side, we are not benefiting as much as we could from that
Orange’s Head of 5G Research, Arnaud Vamparys, said that in fact the operator had a well known commitment to being able to put all network elements on generic hardware – and that the time was now right to extend that to the RAN, to enable network slicing and new edge-based, low latency, use cases.
“Really, on the RAN side, we are not benefiting as much as we could from that, because even though certain interfaces are in 3GPP it does not really go through [in an interoperable way]. We think that it’s the right time to have this interface in the next releases from 3GPP.
“We already have had some initiatives but it was quite fragmented, so we decided to move together to one set of specifications, and then everyone can play with that, new vendors, open source communities – they know software.”
Vamparys said that if you want to step to network slicing then you definitely need to have an open interface to do that correctly. “Everyone can have a black box and sell that, but is it the real innovation? Even if the interfaces are open at the end of the day we can choose an integrated solution but with this interface innovation can appear.”
The Open RAN is the ultimate goal, it’s good, but it still requires feasibility and time to market
SK Telecom’s Minsoo Na, Senior Managed of SKT’s R&D Centre, said that SKT is seriously considering the feasibility and reality of the Open RAN approach. “We are more interested in the open fronthaul interface, by realising that we can get multiple radio unit vendors, even third party vendors. But the modularity concept and software based baseband implementations need more time.
“The Open RAN is the ultimate goal, it’s good, but it still requires feasibility and time to market. We don’t have much time to deploy early 5G network, we don’t have much time to customise or optimise the open software, that’s why we are seriously considering pros and cons.”
Vendors
Just before MWC, Samsung rather took the market by surprise with an announcement that it had been listed by Verizon as a supplier to that vendor’s Open RAN initiative. There were two reasons for this surprise. First, nobody knew Verizon had a formal Open RAN programme. Second, adding a new LTE vendor years after you started rollout is unusual.
The US market has been an opportunity for some of the vendors to fund what’s going on in other markets around the world
Alok Shah, US marketing head at Samsung Networks, said that as Samsung is already an indoor 4G provider to Verizon, and one of its 5G FWA suppliers, adding outdoor LTE made sense. He added, though, that coming in as a new vendor was still a “big deal”, and he rather thought that the announcement should have taken a few more people by surprise.
He also intimated that US operators are getting impatient at being asked to pay top dollar by the incumbent NEPs. “The US market has been an opportunity for some of the vendors to fund what’s going on in other markets around the world. US operators are somewhat frustrated with that”. Therefore operators’ commitment to lowering prices is very real. “We have joined xRAN Forum [now ORAN Alliance] and Verizon is a part of that. There are standardised approaches that we think will have a real impact,” Shah said.
In the initial instance, Samsung will be working to a standardised CPRI interface, so that in some cases it will sell both RRH and BBU to Verizon, but sometimes only its RRHs or BBUs.
Verizon also announced that it had worked with Nokia and Intel on trials to run Nokia’s AirScale base station software as a vRAN instance on COTS hardware. The trial, in Oklahoma City, is a sign that a major incumbent is prising apart its hardware and software in line with operator demands.
Cisco
Cisco announced it was forming something that it termed an open vRAN ecosystem. The vendor wants to act to bring together companies that are active in the Open RAN space – and named Reliance Jio, Altiostar, Aricent, Intel, Mavenir, Phazr, Red Hat, and Tech Mahindra. It wants to act as a facilitator of open interfaces in the vRAN.
We are clearly not getting into the RAN
Emphatically, though, it is not getting into the base station space – at least according to Bob Everson, Global Director, Mobility and 5G.
“We are clearly not getting into the RAN, we are helping integrate a better end-to-end architecture for customers. Operators are looking at a lot of different use cases and deployment models, small cells, enterprise; this might be a time to try out a new architecture.”
Everson says that smaller players cannot bring it all together for operators, just as Cisco “doesn’t want to do this on our own”.
JL Valente, Vice President of Product Management for Cisco’s Cloud and Virtualisation Group, said, “We want to bring new value into the RAN market, a ecosystem to maximise the the potential value of those players. It is not to replace wRAN/ORAN and TIP – we want to take those and put them to work, take those specs and standards and get to a point where it is practical.”
Everson added, “We want to create a reference architecture and start to integrate elements towards that.” That would include interoperability testing and validation, for instance characterisation of transport in band and fronthaul.
JL Valente: “We have seen two to three years of accelerating acceptance of virtualisation, mainly in the core. But if we want to get to the level of flexibility for orchestration and slicing that is required, we need the decomposition and virtualisation of the RAN as well, or at least big chunks of it. Customers need to see virtualisation happening in that area. We see this as the right moment; we’ve been preparing ourselves but incumbents have no vested interest to change the status quo. Meanwhile isolated companies don’t really have the punch to make a big bang in the market. So we are lending the Cisco authority, knowledge and connections to say it’s time that we take a stand for this multi vendor environment.”
So if Cisco is not in the RAN, what is the value to the company of speeding up and validating an open vRAN ecosystem?
Everson said, “There are a number of opportunities whether in transport, infrastructure, orchestration components, mobile edge compute and in the mobile packet core. Our end-to-end view in 5G integrates really nicely.”
So how will operators balance the imperative to invest now in 5G against the desire for the Open RAN. Everson said it would be a case of inserting Open RAN elements into edge cases: “The way we see it is they insert around a particular problem to solve and then expand from there.”
Valente added that the the Open vRAN should not just be seen just in a 5G context.“Not everybody has the pressure of 5G, and it’s something that’s already very compelling for somebody who doesn’t have 5G.”
Mavenir and Adva and BT demo
On Adva Optical’s booth there was a demo of network slicing across a deconstructed RAN, with Mavenir radio equipment and vBBU and vEPC with Control and User Plane separation providing slice control across ADVA demarcation and edge servers. The demo included SDN-controlled front and backhaul with NFV infrastructure for edge hosting. An additional demo had a small cell from Accelleran and a performance assurance toolkit from Spirent. One of the aims for BT is to be able to demonstrate Network Slicing control across backhaul links on a wholesale basis.
THE OPPOSING FORCE
And yet for all the moves to the Open RAN, there exists a dialectic imperative. Open RAN will take time to be commercially viable at scale. But most of the Tier One and indeed Tier 2 operators are under national and commercial pressure to launch 5G, or at least be seen to be doing so. That will necessarily predicate investment with those same incumbents threatened most by the open RAN. And the incumbents know it. So DT’s Jacobfeuerborn may promote the white box RAN, and DT may have drive xRAN and now ORAN, but at the same time when DT announced the vendor to drive its step into 5G, that vendor was Ericsson.
At MWC Nokia stepped up its efforts to promote the benefits of an end-to-end strategy for 5G networks. Its own chipset, designed for a range of different deployment options, Reef Shark is at the core of this. Multi vendor is fine, but end-to-end is much better, CEO Rajeev Suri said.
Ericsson’s CEO Bjore Eckholm all but begged operators to start investing in 5G now. Ericsson found some research to support its claim that operators that deploy new generations early stay ahead for the lifetime of a “G”. Ericsson wants operators to move to 5G as soon as possible because it needs the money. That also has the effect of delaying the move to the Open RAN, and therefore the arrival of RAN new entrants.
The other driver in opening up the RAN has been the support for network slicing across the network, and the move to architect virtual core and radio elements at the network edge. And it is here that we will address the second of our MWC reviews.