Your browser is not supported. Please update it.

8 April 2019

Ericsson and AT&T give network slicing an open source boost

The Linux Foundation’s annual Open Networking Summit (ONS) has become of rising interest to the mobile and telco community as the open source organization has become increasingly focused on telecoms networks. There will be coverage of the highlights in next week’s edition of Wireless Watch, but one development caught our eye even before the event started on Wednesday.

This was a demonstration of network slicing, harnessing the capabilities of the open source ONAP (Open Network Automation Protocol) software, which handles the management and orchestration (MANO) of all the components in a virtualized network. Slicing based on open source orchestration is interesting in its own right, but even more so because the demo was a collaboration between AT&T – the instigator of several major open telecoms initiatives including ONAP – and Ericsson, until recently one of the biggest skeptics of an open movement which could undermine its business model.

The results will be fed back into the ONAP community.

The demo showed different end-to-end slices of a network being virtually carved out to support different levels of quality of service, including a very high QoS slice, which might be for premium customers, or for a mission critical service.

It used live data in a mobile network linking four different remote terminals in AT&T’s labs in Atlanta, Georgia. Two terminals, designated as standard users, ran video streams at speeds of up to 10Mbps – but with the rates falling to below 1Mbps when the network was loaded with large amounts of traffic. At the same time, the other two terminals, designated as premium users, ran video streams too, but their speeds were guaranteed never to fall below 2.5Mbps.

The aim was to demonstrate how availability and performance could be guaranteed even in a network with limited capacity, and in which the air interface had been highly congested. “The purpose is to demonstrate that we can create performance levels of use cases in this network slice,” said Tomas Ageskog, head of digital services with Ericsson North America. “The speeds of the terminals were selected to make sure we are congesting the air interface, which in this proof of concept has been turned all the way down to 5 MHz, so we prove the ONAP demonstration works.”

He hastily added: “It should not be taken as a reference to real speeds.”

Such demoes are highly relevant to the ambitions for E2E slicing – which in 5G can go right from a sliced air interface to the core and transport. Operators hope they will be able to support many revenue-generating services on one network, greatly reducing TCO to enable diverse applications. They also hope they will be able to support mission-critical and SLA (service level agreement)-based offerings, which have been almost impossible to deliver over a mainstream mobile network, but which carry premium fees.

An end-to-end system, which can dynamically assemble resources needed for a particular service on-demand, across wired and wireless networks, is key to this. Ignacio Más, head of technology innovation in OSS at Ericsson, said the demo showed how, in future, operators could connect the different QoS mechanisms of different domains, with different configurations, into one single end-to-end QoS mechanism.

In the demo, the physical radio and virtualized core network elements were configured automatically using ONAP, which also orchestrated the virtual network elements required to form the slices. The slices were created with a single click that executed a multi-tiered workflow automatically.

“This has been built on only the second release of ONAP – Beijing,” said Ageskog. “The key thing is the team has been able to prove that this works. The way that we initiate the service is managed in one step. That’s quite impressive.”

The participants said the demo went a step further than most other slicing events because it showed how the system could instantiate virtual functions; configure both virtual and physical functions; and then combine them to create service-specific slices with unique configurations and QoS settings.

Clearly, it is this ability to combine and re-combine a huge variety of elements on the fly, to support the requirements of virtually any service or user, that is the vision for slicing, but that will be a tough one to turn into reality. Most of what is currently labelled slicing is really an extension of virtual private network (VPN), or affects only the fixed access network. Sometimes it seems a long road between that limited definition and the fully dynamic, automated end game, and most operators still talk mainly in terms of a few large, static slices to support a particular industry, such as automotive, or a particular service, such as high quality video.

But demonstrations like this one give a glimpse of how the vision can be turned into something commercial, and it is interesting how much of that work is being driven from the open source community, albeit with significant impetus from major corporations, even those with much to lose from a full shift to openness, like Ericsson.

One of the obvious qualms that operators have about open source initiatives is the fear that they will not be commercially robust, or that it will take significant time and cost to turn the open source platform into something carrier-grade.

The Linux Foundation was keen to stress the telco-class nature of what it is hosting, and said that the demo had set out to prove the robustness of the infrastructure, and the ONAP elements, as well as to show off slicing advances. Issues such as the ability to incorporate existing components such as base stations into the slices alongside the virtual functions are not unique to this issue – they run through all the work that the Foundation is supporting on virtualizing telecoms networks to support open, software-based architectures and resultant new business models.

Ericsson seemed to go through something of a conversion to open source a couple of years ago when it became a platinum board member of the OpenStack Foundation and joined the ONAP Alliance. More recently, it even joined the ORAN Alliance, another LF-hosted, AT&T-initiated group, which is looking to define open interfaces and architectures for an open, disaggregated RAN.

The LF Network Fund:

At the start of 2018, the Linux Foundation announced it would combine six major networking and telecoms projects under a common “horizontal umbrella” called the LF Network Fund (LFN). This has similar objectives to the newer LF Edge and Deep Learning Foundation, and the older Cloud-Native Foundation – to create common frameworks which will encourage interoperability, and therefore confidence to deploy.

That means rationalizing a host of activities in all these key areas of technology, to reduce fragmentation and market confusion. This could also mitigate the risk that operators will lack the confidence, despite the allure of open platforms, to tear themselves away from their familiar vendors, standards bodies and closed platforms. It will also help to reduce the variety of skills and resources telcos will need to support all the projects, and LF will charge a single fee to belong to LFN.

The LFN will bring together six projects at various stages of maturity. They are:

  • ONAP (Open Network Automation Protocol)
  • OPNFV (Open Platform for Network Functions Virtualization)
  • OpenDaylight
  • io (Fast Data.input/output)
  • PNDA (Platform for Network Data Analytics)
  • SNAS (Streaming Network Analytics System)

Individual projects retain their independence but the Foundation provides mechanisms for communication and sharing, and enables “avenues for greater collaboration” between them. In LFN, this is achieved through two bodies, a technical advisory council and marketing advisory council.

Arpit Joshipura, general manager of networking and orchestration for the Linux Foundation, said last year: “Over the last five years, open source has picked up steam significantly and as specific use cases or problems arose, multiple single projects were created, and they have been deployed and have had momentum. Now there is an architectural discussion going on across projects and there is a need for looking at all these components from a solution perspective.”