About 13 years ago I received a tip from some Silicon Valley insiders that I should check out this “hot” new startup, Timetra, which was blazing new trails in edge routing. That’s when I first met Basil Alwan. Timetra was a pioneer in a niche product known as a services edge router, a device that could not only perform standard routing functions but also added flexiblity with Quality of Service (QoS), using the MPLS protocol to slice-and-dice various IP-based communications services and assign them different levels of priority. This is an approach known in the business as “tag” or “label” routing.
Several things impressed me at the time. It wasn’t just Timetra’s solid technology — which soon made it one of the most successful edge routing companies of that generation — but Alwan himself. Alwan wasn’t your typical chest-thumping Silicon Valley entrepreneur. He was a non-nonsense “doer” with a soft-spoken manner. I guess what I’m saying is that, for the Valley, he’s a refreshingly humble and nice guy.
Alcatel acquired Timetra in 2003 for the bargain price of $150 million (and subsequently completed a gigantic merger with Lucent in 2006). Nearly 12 years later, Alwan is still at the combined Alcatel-Lucent, leading its IP division. Timetra has become the engine of Alcatel-Lucent’s entire IP routing strategy. Many experts have observed that Alwan’s Timetra crew and its technology have had an outsized impact on Alcatel’s organization, helping build its IP routing company to become a formidable rival to Cisco and Juniper.
After more than a decade, it was time to catch up with Alwan again. He agreed to an interview, which we conducted about three weeks ago. At the top of my list was to find out what made Alwan stay on so long after an acquisition. Alwan shared his thoughts on this as well as Timetra’s great run at Alcatel and what the future means for network and service-provider technology.
Rayno Report: So you are a startup guy that has stuck around a huge company more than 10 years after the acquisition. That almost never happens. What’s made you stay?
Basil Alwan: It’s not just me, most of the Timetra team is still intact. Although it’s happened a few times, it is unique, and it doesn’t happen all that often. When a startup gets acquired, there are several things: There’s financial success but then there is market success. Some small percentage of startups, maybe 20%, have financial success, but then of that percentage, a much smaller percentage has commercial success.
I’ve always thought that the companies having commercial success are the most interesting. The best companies in the Valley are about teams that have commercial success in the market. It’s about contributing in a big way and liking what you do.
RR: So, it’s really rare.
Alwan: Yes, of the 20% [of startups] that are successfully acquired, maybe 10% to 20% of them have meaningful commercial success. These are rare events — they don’t happen that often.
When a team is really good, the people are there for the right reason. You still have to have the structure of the team in place to keep it a success. You have to be careful in large companies so that politics don’t get crazy.
RR: Plus, for the employees, it’s not that easy to find a stable, comfortable job in a successful organization these days.
Alwan: Absolutely. Our engineers get to do challenging things, too. We’ve had secular growth for 10 years running. It’s something you always think about: Do you want to do other things, or should you stick with it?
RR: These days, everything is about cloud. Cloud started outside the service provider world, so what does cloud mean for service providers?
Alwan: Cloud has been around awhile, and applications have been in the cloud for awhile. What’s created the situation is the access buildout, which the service providers did. The access bandwidth enabled the cloud.
I did a presentation at our analyst meeting. We calculated how much was spent in wireline access in North America — we concluded that 1.8 petabits was added to access. IP has become the lingua franca. There are three things that have contributed to this: 1) Huge investment in access over a decade; 2) Migration from many network technologies to IP as the common platform for applications; 3) Evolution of compute via Moore’s Law has made economics of shared compute even more compelling. It’s created this situation where applications want to move and run wherever they want to run.
At the high level, the networking industry is a relatively static network — expecting resources to be in specific locations, ever expanding capacity. We’re still doing that. The peak capacity evolution at a major service provider is 50 to 75%. The new thing we have to think about now is driven by cloud. We need a logical network on a physical infrastructure.
Even if you don’t offer compute in the cloud, you have to think about how well it works with the cloud, which means it’s much more agile. Our customers are having to think about it. On top of that, they have to think about how to offer double the capacity.
RR: What does it mean for the equipment guys?
Alwan: Everybody feels like they have to market their stuff as Software Defined Networking [SDN]. A classic network management system is not in the spirit of what SDN is about. That’s why we created Nuage [Nuage is a separate division of Alcatel-Lucent]. It’s about making technology from the ground up that is hardware independent. We are really pushing the limits of reconfigurability. This is where SDN is really having an impact.
RR: You are talking about a box industry that is trained to sell boxes.
Alwan: Yes. But there’s a good model for selling software. As part of the platform, all the applications run on the 7750 [Alcatel 7750 Services Router] and all of them run as Network Functions Virtualization [NFV]. So if you want to buy a hardware router, you can, or you can buy software running [on commodity hardware]. When you bring them up they behave the same.
If I’m a customer, I want a function. We have 20-plus applications. What we’ll do is we’ll say: Buy either [the hardware or the software version].
For these functions, that’s a shift, but it’s not as big a shift for Alcatel-Lucent as you think. Whether it’s billing or network management, there are things we sell on off-the-shelf hardware. Even if you are a company that has only built boxes and you’ve never sold software, you can still learn.
RR: How many of these functions can go to commodity hardware? What about the argument that high-performance routing will never go all the way to commercial silicon?
Alwan: It comes down to the fact that packet processing is very similar to DSP [Digital Signal Processing]. Can you do all the packet processing you can do in a router in merchant silicon? Yes, you can. Can you do all the DSP? Yes, you can. Does it get to the equivalent price/performance that you shouldn’t make it purpose-specific? The answer is no, the delta is still quite large.
In very large data-plane applications with a lot of traffic, for the foreseeable future, you are going to have dedicated systems. You can do lots of applications that you run in routers. That routing layer will still be there, and it will be incredibly important that it’s stable. Some of it we will sell as NFV.
RR: So, you are saying it’s an evolution.
Alwan: Don’t get me wrong, it’s a big change. Our customer networks are evolving organisms. They are going to take steps. With the next-generation gateway or edge router, are we going to do it in the same way? No. Broadband gateway, or mobile edge, or business service edge will evolve each time. Over time there will be more NFV functions. That’s the conversation going on right now.
RR: What percentage of customers are buying your network functions as software, and how fast is it changing?
Alwan: Right now, it’s very little. Some things will move faster. For example: A policy manager or an MME [Mobility Management Element], a device that is only doing signaling, are good candidates. An MME is a pure compute device, so it is a great candidate to become a Virtual Network Function [VNF]. These things are not carrying the data traffic, so why not do it as NFV? IMS [IP Multimedia Subsystem] is another example.
The question and answer is not NFV or not — it’s both. It depends on the function you are talking about.
RR: How important is service orchestration to this next-generation model?
Alwan: At a very high level, SDN is a very big factor to run service orchestration because it abstracts services from the hardware. So, it’s super important. A lot of it depends on how the services are being set up and how it’s being set up.
Then there’s cloud SDN, where we’re abstracting all of the services information out of the hardware. Datacenter is a great example of this — they are trying to do services management to instantiate the service in our switches. Or you can virtualize it and put it in the servers. Service orchestration is a great topic, and it’s hugely important.
RR: What should we expect next and in 2015?
Alwan: I was talking to a customer last night, and we were talking about the industry. We’ve always said we were flying a plane and rebuilding it at the same time. Now we’re flying the plane, rebuilding it, and replacing the staff and pilots at the same time.
We have always had to provide more capacity, and now we have to provide an agile way for the services to be abstracted and added to the network. The foundational pieces are around, but the big project in 2015 is making these real deployments. We will have real deployments of NFV and SDN implementations. 2015 and 2016 are going to be the years when we really get NFV and SDN in real large-scale implementations.
RR: Thank you, Basil