Sanqi Li, Huawei

Sanqi Li, Huawei

The CTO of Chinese equipment vendor Huawei clearly envisions a world where network applications and services will run on commodity hardware and software. For a company that sells specialist boxes into networks, this could be a bold statement, but Sanqi Li, CTO of the firm’s carrier network business, simply sees it as the evolution of business models for both carriers and vendors.

Speaking about what he calls the “new digital economy” Sanqi believes the next generation of technology, particularly virtualisation, will bring about more change than the industry has seen in the last 100 years, especially in terms of business models. While he acknowledges that the operator is already the foundation stone of all connections, he argues that operators need to move beyond what is traditionally perceived as their position in the digital ecosystem.

“This next change is less about technology and more about business. The operator business model at present is rigid, slow, and hard to change. Network architecture today is closed, complicated, and still focused on elements that the carriers can control,” Sanqi says, alluding to the idea that telcos should be partnering with OTT services using technologies developed by the vendors. Indeed, with an assertiveness that some of his competitors might be wary of displaying, he adds: “Vendors are increasingly responsible for the operator business model.“

Sanqi believes that the existing TCO structure for network operators is not sustainable, because opex is still growing and is now at 82 per cent, up from 77 per cent ten years ago.

“We’re looking at the fundamental limit of existing network infrastructure,” he says. “Today’s network architecture is closed, protocol-centric and vertical, fragmented and with software embedded in the network devices.” The devices might be in the network for ten, 15 or even 20 years and all services are best effort with very little differentiation between them, he adds. Networks are over complicated by the continuing presence of two decades’ worth of technology, from TDM to ATM,  and frame relay to IP and MPLS.

“These networks are multi-generation and multi-technology, and control is always at the centre of the model. Voice and messaging are all controlled directly by the operator, but this traditional service model cannot be sustained because it’s slow to adapt,” he warns.

A business model that is rigid, slow, and hard to change, can be saved by the core virtualisation expectations of SDN (software defined networking), he says. He characterises the move to this new architecture as a move from closed to open; rather than having a network that scales up, it’s about having one that scales out, across different sectors and verticals, while enabling new network capabilities.

“It’s about stopping the operators from having a controlling mentality and giving them an enabling mentality. Then the telcos’ true capabilities in network, QoS and billing can really come into their own,” Sanqi says.

Huawei, which caters to cloud service providers as much as it does traditional telcos, is talking about bringing cloud business technology into the telco network, using SDN to dynamically allocate resources and spectrum across 2G, 3G, and 4G as well as to provide a virtual gateway direct to the customer’s home.

In terms of the core technologies, Sanqi talks about Network OS, which decouples the control layer from network devices, as well as Cloud OS, which deals with massively distributed datacentres, virtualisation and automation. The combination of these two OSes, touches every part of the network and structurally optimises it.

Sanqi reckons that decoupling, or ‘abstracting’ the control layer can reduce the time and effort needed to manage network devices by a factor of between five and 20. “Abstraction in the cloud layer can lead to simplicity, where providers can automate and orchestrate new services and apply different cost strategies,” he says.

It’s a big shift for an industry that has historically built networks around specialised appliances and one that will need critical mass from the vendor community before the carriers buy into the idea wholeheartedly. But Sanqi says that this approach does not mean moving everything to x86 hardware and argues that it would not have a detrimental effect on Huawei’s business as a maker of both network hardware and software. SDN and virtualisation means network functions are no longer embedded in devices, with control becoming more of a software-based model, breaking away from a costly status quo in which controllers for many remote devices are run vertically, he says.

“More and more hardware shipments will be focused on data centre boxes, the likes of which Huawei sells. These are not just x86 processor boxes, but those that might also require GPU power for video rendering,” he says.

In terms of architecture and software, OpenFlow is expected to be the chosen switching specification, while the Cloud OS will be based on OpenStack. Huawei is building a centralised routing control algorithm and the network OS is still to be decided on, but a key feature is that it decouples control from the devices themselves. The company already has SDN-enabled routers deployed alongside OpenFlow-tested equipment, Sanqi says.

Moreover, an increasing proportion of Huawei’s future product development will be in software, as network functions, no longer embedded in specialist network devices, will all be controlled by software that needs to be open, abstract, programmable and virtualised. How long this process will take is anybody’s guess, though, as represents a move to an entirely new business model.

“Carriers won’t swap everything out at once. This is more of a migration and needs to be managed carefully as it can touch all parts of the system. But the endgame is that carriers are looking for programmability in software to open up the capability to work with OTT,” he says.


Comments Post a comment
  • [full disclosure: I work at a vendor in the SDN space]

    I think the idea that SDN is a business shift as much as a technology shift is right (even if it does underestimate the profound technology changes that are afoot). I think many people have assumed the provider space will be provider vs. OTT. And in that world, the prevailing thought is that the providers are relegated to providing public infrastructure (much like roads). This is likely an unsupportable model long-term, so some have predicted a dire outcome for the carriers.

    However, the network has information the OTTs could use. And this does mean that the competitive landscape could be provider/OTT vs. provider/OTT. Imagine Facebook teaming with Verizon to offer streaming movies to groups, leveraging their content relationships with Warner Bros (for example).

    For this model to work, though, there has to be some means of exchanging information between the network and the applications the OTTs are building. This is where SDN provides the most substantial value IMO. The view that SDN is about making cheaper hardware or automating some network tasks is myopic. There are bigger things afoot.

    So the question is: what data models or abstractions are going to be useful in creating the bi-directional exchange of information between network and service? That’s where the technical discussion ought to gravitate.

    Reply to Michael Bushong (@mbushong) on SDN: More new business model than new technology
Post your comment

@telecoms