Neil McRae, BT: Driving force

11 August 2014 | Guy Matthews

He's never happier than when he's got his hands covered in motor oil. So what is Neil McRae going to make of a clean, well-ordered world where everything is seamlessly driven by software? Guy Matthews interviews BT's chief network architect.

Neil McRae, chief network architect with UK fixed-line incumbent BT, is not a man who appreciates technology for technology's sake.

"I'm sure it's a frustration for my marketing colleagues, but my approach is always to ask 'What challenge is it that we are trying to solve?'" he says. "You don't want to just have a cure looking for a disease. When I hear people talking about a technology that will change the world I think 'Hang on a minute'. The technology that survives and prospers is the one that solves real problems and adds real value."

So what does McRae make of software-defined networking at this early and partially formed stage of its development, surrounded as it is on all sides by frothing enthusiasm and heightened expectation?

"The whole SDN and NFV thing is in a hype bubble, to be honest," he believes. "It's going crazy right now, with a lot of vendors looking at the situation as an opportunity to sell things they've already got with a new label on them. 'Mmm, I think I can market this thing in a different way'."

Two sides to every story
It would be quite wrong, however, to paint McRae as a full-blown SDN cynic. The technology's plus points are not lost on him: "SDN is a great way to orchestrate and manage data centre applications and their connectivity," he enthuses. "Cloud services can be set up in minutes, but without SDN the networking of them takes so much longer. SDN gives you the opportunity to have a system whereby when an application is created, it can signal for the connectivity services it needs. There's value in that."

McRae points to the obvious gulf in SDN achievement between WAN and data centre: "That's where SDN is from," he says. "If you're a Facebook or a Google then it's vitally important for you. There's lots of virtual infrastructure you need to be in control of, and data centres all over the world. But it's not the same if you're operating a fixed network. Big data centre? Fine. Fixed-line operator? A lot to sort out."

He sees a future for SDN at WAN level as a way to orchestrate applications and add new services, building them and pushing them out to the network.

"It means you don't have to manually log in to a lot of different boxes," he says. "You can be challenged by all the different firewalls when you are setting a service up. In an SDN world, it lets the firewalls know that something is approved."

A prime use case for SDN today, in McRae's opinion, would be for a mobile operator rolling out something new and seeking to shave off some of the time it takes to do so, in a tight market where any lag can be fatal to the chances of a newly launched product.

"MNOs tend in any case to be more IP-driven, rather than driven by infrastructure," he points out. "They can deploy the power of SDN to enable and sell new services in areas like high-definition voice, for example. Service deployment for mobile operators can be very challenging. I can see that SDN would have made a difference to the development of MVNOs if it had happened five or so years ago. That would maybe have been a different story."

He's not so certain that SDN at this stage can add anything more than peripheral value for an owner and operator of fixed network assets: "BT's broadband platform is already heavily automated," he explains. "Is SDN ever going to make a massive difference to it? Probably not. We manage millions of users, so we've hugely invested in ways to do that."

BT will be using SDN for certain purposes, he says: "It will be useful to orchestrate BT Cloud Compute, and we'll use it along with NFV to manage equipment and services at the customer premises. We'll use it also to help content distribution on our 21CN infrastructure. Those are the three main uses I'm seeing. And I see it in those areas more as an evolution and not a revolution. The idea that challenges will just melt away with the power of software is just not the case. We'll use it where it makes sense."

For a new operator building from scratch, he believes SDN is a more justifiable and sensible approach. But he warns against the romantic notion that the right software suddenly does away with the need for all physical and human aspects of a network even there.

"SDN is a control mechanism, but you still need things like copper, fibre, exchanges, offices, the guys in the NOC," he says. "SDN is not some sort of revolution that will just sweep all that away."

Big youth
McRae is worried when he sees some of the expectations around SDN matched against its extreme youth: "It's still in the very early stages of development," he warns. "You have to have in mind the Gartner maturity curve – where early hype gives way to panic when people can't make a thing work. They run away, then eventually they come back and figure it out. We are very early in that cycle."

SDN's callow youth is not just evident at carrier level, but in the R&D activities of the vendor community too, in McRae's view: "There aren't actually that many vendors that have launched SDN platforms yet, it's that immature a technology," he believes. "We've seen something from VMWare, Alcatel-Lucent with Nuage, Cyan and Cisco. Though people who talk about 'vendor lock-in' probably just aren't managing their vendors right."

McRae has had recent face time with a number of the big SDN vendors following a trip to California: "I talked there to Cisco, Juniper and also a lot of startups, too," he says. "I felt a bit worried for them. By the time all their work has been done the technology may already be commoditised. It just makes me think that a lot of vendor R&D might have been better employed in other areas of network development – like getting greater throughput of data through fibre or caching efficiency."

McRae believes that there are many vendors trying to build an SDN vision, but with too little unanimity between them: "It's a bit of a land grab at the moment," he fears. "But I don't know that these visions will necessarily work for BT, or for a large enterprise user for that matter. We're just too heterogeneous. The returns that can be demonstrated on any user case so far are quite marginal."

High levels of expectation for new ideas are nothing unusual in the communications industry, of course. But McRae thinks the excitement that surrounds SDN is different: "There are previous big hypes that we've seen in this industry," he says. "But it was obvious in the early stages that, say, MPLS and virtualisation would make a big impact. It's less obvious with SDN."

SDN, says McRae, is often touted for its low development costs. But he argues that while it can cost a lot to develop hardware, good carrier-class software is expensive too: "Juniper Networks has got whole armies at work on this," he says. "It needs to be very good software too. We have a lot of mission-critical stuff that we run, systems for healthcare as well as content for TV. We can't just have that switch off on us. It's often the case with software that as you solve one problem you generate another problem elsewhere. You also have to question the real benefits. Quicker provisioning is useful, but it's not going to make our P&L jump over the moon."

He is also anxious that having too much pressurising expectation around SDN could confuse enterprise CIOs, unless the vendor community is careful: "Too much hype gets them worrying what it is they need to rush and do," he fears. "But there will be an SDN upside for them – over time."

McRae is not alone in believing that the noise around SDN is possibly masking a lack of standardisation driving the technology. Much work undoubtedly remains to be done.

"I like OpenDaylight – I think it's a brilliant initiative where they are not just talking, they are doing the work," he enthuses. "But with standards in general, I think you need to consider the early days of the internet 20 years ago. Some of the standards then took a long time to get together. Some are only just maturing now. We're not seeing that with SDN right now. There's three or four approaches to the same thing wherever you look, some of which verge on the proprietary."

Too many cooks
McRae thinks that, nonetheless, much good standards work is going on: "It's just that a lot of people are trying to do the same thing in different ways, and there's a lot of vested interests too," he says. "Everyone wants different things to be incorporated into the standard according to what their priority is. One thing we can't afford much of at this stage is duplication of effort."

McRae has seen his share of new ideas come and go, some sticking, some fading away. With his specialised focus on IP and internet networking and applications, he has gained industry-leading knowledge and strategic insight over the years into optical WDM and transmission, VoIP, FMC, xDSL, Ethernet, OSS/BSS, and web applications technologies. In March 2013, he received BT's SMT Award from company CEO Ian Livingston, for his leadership in network transformation.

So what of his off-duty hours? "I'm afraid I'm a bit of a petrol head," he concedes. "I like doing crazy Top Gear challenge stuff, like building stuff and making it go fast. I've been involved with the building of a diesel car that can beat a Ferrari, and driven across the US and all over the Alps. I'm most happy when I've got my hands covered in oil."