Your browser is not supported. Please update it.

22 April 2016

Is open source the only road for NFV?

Every vendor and operator seeking to play in the new world of virtualization and software -defined networking (SDN) must claim to be ‘open’. One of the driving motives for carriers to move towards a software-dominated environment is to escape the old proprietary platforms and lock-ins. However, how different groups achieve ‘openness’ varies widely, and threatens the same kind of fragmentation and dead ends that afflicted deployers of proprietary hardware-focused networks.

There is often confusion between openness – as in interoperability between different vendors’ solutions, enabled by standard interfaces and APIs – and open source. But increasingly, carriers see the two as inter-dependent, and the way to accelerate progress towards commercial SDN and NFV (Network Functions Virtualization) solutions.

In a recent interview, Verizon’s director of SDN/NFV network architecture planning, Gagan Puranik, told RCRWireless: “We believe the pace of SDN and NFV development could be more aggressive. We are aggressively driving open source/open interface initiatives to enable large ecosystems to accelerate progress with focused project initiatives.” He added that the key to speeding up the process was “open standards and open source working closely, with a sense of urgency”.

Even ETSI, the body behind the core NFV specifications, began work this spring on an open source implementation of NFV’s management and orchestration (MANO) components, saying: “Open source software can facilitate the implementation of an ETSI-aligned NFV architecture, provide practical and essential feedback to the ETSI NFV ISG and increase the likelihood of interoperability among NFV implementations.”

This move illustrates the fact that open source may increase fragmentation rather than alleviate it – there are at least nine open source SDN controllers, for instance. The approach certainly spurs innovation and a wide ecosystem, but rival open source platforms will need to be converged under the auspices of a formal standards process too, to ensure long term openness in the most real sense.

ETSI’s Open Source MANO (OSM) project is challenged by other implementations of NFV which rely on OpenStack for managing and orchestrating all the virtual network functions (VNFs).

Some early triallists of OpenStack NFV have been critical of it, arguing that it prioritizes simplicity over functionality. Last year, BT’s chief data networks researcher, Peter Willis, told a conference that the firm was considering proprietary alternatives to the technology for its virtual enterprise CPE roll-out, unless a set of issues was addressed. These related to the connection of VNFs, service chain modification, scalability, ‘start-up storms’, security and backwards compatibility. “If these six issues are not addressed we will not use OpenStack for virtual enterprise,” Willis said.

Indeed, OpenStack does not carry out all the MANO functions as defined by ETSI, though it could evolve to do so. Critics say OpenStack is just a virtual infrastructure manager (VIM) – one among many cloud management systems which may be chosen by operators to host NFV – whereas ETSI MANO would deploy VNFs above the VIM for greater platform flexibility. Opponents also say that OpenStack is inadequate for managing legacy networks alongside NFV, as most carriers will want to do.

On the other hand, supporters of OpenStack say it is a sufficient option for NFV and has the advantages of relative simplicity to deploy, accelerating time to market, and of a wide ecosystem. A recent report from the OpenStack Foundation said NFV was the fastest growing use case for the technology, something CEO Jonathan Bryce had not anticipated. “It turns out that some of the telcos wanted to automate their resources. They came to our community because they thought OpenStack had the potential to help them do that,” he said in an interview.

He told Computer Weekly: “Firstly, cloud is the right way to manage resources dynamically and they need to do that for their networks. In our case the existence of the community was also key. The big driver for NFV is to lower cost models. Participating in open source communities is very disruptive to entrenched business models that telcos have followed for decades.”

Because of its wider roots, OpenStack can orchestrate many non-NFV elements within a virtualized or cloud environment, helping to integrate the new networks with data centers, uniting NFV and the cloud.

The latter point is an important one for the large vendors, which are approaching NFV either from a starting point in the carrier network, like Ericsson, Nokia and Huawei, or from the data center and enterprise network, like Hewlett-Packard Enterprise (HPE), Cisco and Juniper. Both sets of suppliers are eager to bridge the cloud and networking worlds in order to offer their customers an end-to-end platform and entrench themselves in the organization.  These large vendors are including OpenStack in their portfolios, and some are looking to increase their market reach by open sourcing their efforts too, as seen in the open version of Juniper’s Contrail. HPE recently made it a requirement for any supplier in its Cloud 28+ catalog to be OpenStack compliant within a year, as the simplest way to guarantee interoperability, allowing operators to spin up a Cisco or Brocade router on commodity Linux hardware, without being reliant on their own chassis.

And OpenStack is certainly building up a head of steam among operators. AT&T was the first to use the technology commercially for NFV and now runs several data centers ni this way, and is in the process of virtualizing a large proportion of its network functions over the next few years. Tele2 recently turned to Canonical for an NFV-compliant virtual EPC (evolved packet core) deployment; while Japan’s NTT Docomo boasted of achieving the world’s first interoperable multivendor NFV platform, with Ericsson’s OpenStack-based Cloud Execution Environment holding all the elements together. And SK Telecom has OpenStack as a key element of its experimental 5G networks.

No wonder, then, that the vendors are rushing to embrace OpenStack and go through the painful process of ripping out their proprietary network roots. Nokia, fresh from acquiring Alcatel-Lucent with its rich cloud and SDN initiatives, has unveiled a new generation of ALU’s Cloudband platform, boasting convergence with Nokia’s own offerings and a new level of openness. The new CloudBand, available commercially in July, will allow operators to mix and match products from other vendors, and Ron Haberman, VP of the CloudBand business unit, said in an interview: “Not only do we now have well delineated layering compatible with ETSI, but quite a bit more focus on open source in general.”

So CloudBand 4.0 is promising a flexible, configurable OpenStack portfolio in which competitive VNFs can be implemented with common orchestration. And Haberman believes it is important to be open source as well as open, which means participating in a full community – “It’s not enough to simply take the code and publish it in a public repository to be considered open source, in my opinion,” he told FierceWireless.

Nokia, like its primary rivals, does not just want to ship technology, but to become the integrator of the operator’s whole complex SDN/NFV platform. CloudBand is designed to provide integration and visibility across every layer of the network, and on that foundation, Nokia will provide tools and services to support a range of specific use cases, many of them focused on mobile networks (especially VoLTE) and the Internet of Things.

“There is visibility to the operators of each layer so they can see what is happening in the other domains,” said Haberman. “And we have the ability to operationalize a particular use case. That is how we see most operators moving forward, on a use case basis.”

CloudBand 4.0 centers on the former ALU offerings, CloudBand Node and Manager, which suport NFV infrastructure, resource management and MANO functions. It also incorporates VNFs and APIs developed at Nokia but retains its original three-layer structure – the CloudBand Infrastructure VIM, the Application Manager for open VNF management, and the Network Director for orchestration and service chaining. Nokia said it will introduce a certification program for third party VNFs, and is making carrier-grade additions to the OpenStack platform and contributing those back to the open source project.
The latter effort shows how vendors acknowledge the limitations of OpenStack, but still see it as an open, ready-to-use framework on which they can build the additional functions and robustness which carriers need. Haberman believes the creation of a truly carrier-class stack will be Nokia’s chief competitive edge in this market, working with partners such as Red Hat, which has just announced general availability of its OpenStack Platform 8 with associated Cloud Suite, promising automated updates and several of the features carriers are demanding. The additional tools which operators need include automated installation, in-service upgrades, various packet acceleration techniques and others, and Red Hat says it is previewing various tools to improve NFV, including

more predictable latency with real-time KVM, improved network I/O performance with DPDK-accelerated Open vSwitch; and an OpenDaylight networking plug-in to support SDN.