It has taken a long time for the open source revolution to reach the mobile operators. 15 years ago, the enterprise data center market had already embraced this approach, and vendors from IBM to Red Hat had adapted to it or created new business models around it. Only now that the MNOs so urgently need to transform their cost base are they, too, pushing for a new open ecosystem, and the vendors are reluctantly having to start on their own transitions.
It remains to be seen whether any of the big OEMs manage to do an IBM and turn open source into a weapon rather than a scourge – or whether, as the major operators darkly hint, the network will become the preserve of emerging start-ups which were founded on a more open model, and on differentiation through software not hardware.
Certainly, no supplier can run against the tide. The 5G network will end up as a mash-up of open standards, open source and proprietary technologies, but any company which wants to play will have to offer commoditized white boxes, interoperability and very clever software. In some cases, differentiation will come from adding new layers, or services, to an open source stack; in others, more traditional standards licensing will continue. In a decade’s time, however, the cost of deploying a mobile network will have been radically altered.
Hardware’s share of network revenues to fall by 50% in five years
The powerhouse behind the open source aspect of this is the Linux Foundation, which held its Open Networking Summit last week. To mark the occasion, it published a white paper, written by ACG Research, which detailed a “massive shift” to open source within the telco industry. This market will be worth $11bn by 2023, estimates ACG, much of that revenue coming from management and orchestration (MANO) software, and from edge compute.
Both these areas are highly strategic because they are essential to the promises of virtualized end-to-end networks and of new economics for 5G. Both will have open source underpinnings, but on top of those, vendors can create optimized and differentiated added value solutions and tools. The MANO software sector will double in size by 2023, and edge networking will more than double each year in the same period.
Despite the software focus of the new networks, however, $7bn of the total will still be in hardware, mainly for the edge. That is 63% of the total across the five years, whereas the figure in 2017 was 72%. At that rate, the figure could be expected to be close to 50% by 2023, or even less if white box routers really take root (see separate item). The report expects 35% of systems to be using merchant silicon rather than proprietary ASIC chips by 2023, with the former reaching even into the high performance extremes of top end switch-chips, as highlighted by launches from Broadcom and Barefoot to power white box switches. In these environments “the vendor/brand value will be locked in the network operating system (NOS),” said the Foundation – and some organizations, like AT&T with its dNOS technology, will look to open source that system too.
Large sums, then, will still be spent on the new networks, but these will be considerably lower, on a per-subscriber or per-Gbps basis, than traditional architectures. More questionable is whether the open networks will push down opex too, since most operators expect to spend more on inhouse skills, or external consultants, to deploy and optimize the modern platforms.
However, over time it will spark a change in processes and should foster more collaboration between operators, suppliers and partners, leading to a more fluid supply chain and sharing of development efforts and risks. “Open source is a critical driver for change and the catalyst for digital transformation underway at many operators,” the white paper said. “It offers compelling benefits to both operators and vendors – perhaps most importantly, the collaborative innovation that only a diverse community can provide.”
These advantages, in addition to cost benefits, will make open source “the default choice for operators” for network MANO, commented Arpit Joshipura, general manager of networking and orchestration at Linux Foundation. “Vendor partnerships and embracing open source will be part of the solution; those who stay proprietary will have a challenge. That is the key takeaway.”
One operator which is embracing open source orchestration is Telefónica. The operator is introducing ETSI’s Open Source MANO (OSM) into the heart of its massive Unica virtualization program. The Spanish telco contributed a lot of the code for OSM, but has not, until now, considered the system sufficiently robust to be included in Unica. It has initiated the process of selecting an integrator for its OSM deployment, highlighting how telco software firms are seeing new revenue opportunities in open source integration services. Amdocs with its engagement in ONAP (see separate item), and Radisys through involvement in CORD are both good examples of this shift from software products to deployment services – a shift that was also the secret to survival for many data center software vendors in the past too.
Telefónica has been onboarding virtual network functions (VNFs) with OSM since January 2017 and now needs to integrate OSM with OpenStack, which underpins Unica. This will be the key role of the integrator, and the project, if successful, should boost confidence in OSM against its main rival, ONAP (Open Network Automation Platform), which is hosted by the Linux Foundation.
The appointment of an OSM integrator reveals that many open source projects are coming to commercial maturity, but also that this takes almost as long as with traditional platforms (longer, if there is significant fragmentation to address and resolve).
For most, the solution will be judicious mixing of both elements, and the critical enabler is a set of open interfaces to allow that. “Open source is the way we think we can move faster, with some challenges, of course, in making all these developments carrier grade,” said Juan Carlos Garcia, Telefónica’s director of technology and architecture. “We have to complement the gaps in open source with industrial solutions and supplement these open source projects. We need a simple way to integrate all this stuff and we think a common information model will minimize integration efforts in the future.”
There are still hopes, following reported overtures by Telefónica to ONAP last year, that the two platforms may converge, or at least work on interoperability, to avoid a split in the industry at this early, critical stage for virtualized telco networks. A common information model between OSM and ONAP could be the smoothest way to achieve this.