Today we’re going to dig down into layer 1, the transport layer that underlies the entire sector and actually carries all those bits from place to place around the world. XKL is out to reinvent the way operators and enterprises use and implement dense wavelength division multiplexing in their networks. With us today to tell us how XKL approaches the optical transport market and what they are seeing in the marketplace going forward is Chad Lamb, Director of Engineering at XKL.
TR: How did XKL get started, and how did you get involved?
CL: Len Bosack, co-founder of Cisco Systems, started XKL in the early 90’s after leaving Cisco. He’s not the kind of guy who sits on the beach in the Caribbean, he wants to get things done and improve the network. He started XKL working on high-speed computer architecture and some side projects that he’s always wanted to work on. Basically, he built XKL from the ground-up to what it is today. Soon after that, he started working on the products we work on today – optical networking equipment.
I've been with the company since the late 90s – I came from academia and had a particular interest in hardware. I started as a hardware engineer and through the years I've worked my way through the organization to my current position leading the engineering team.
TR: How does XKL approach optical transport, and how does it differ from the larger, more established vendors?
CL: XKL got started late in the optical networking and transport area. The benefits of that are that you can see what's in the industry and talk to customers about what they like and dislike about the equipment and software. Then you can build something from the ground up that has much more customer feedback put into it. After a few years, we had a really good sense as to what worked and what didn’t based on this feedback. We were able to take that into the early stages of the design cycle and build a product that is really customer driven. The basis for the XKL optical transport equipment starts from those initial conversations with the customer. It’s unlike other solutions where they started a product in a niche market with one customer and grew it from there. We were able to start from a fundamental level on what was really needed. What you see in our product line today is the culmination of that, both in hardware and software. That translates into a better product, at a better price, with the feature set customers are looking for.
TR: So XKL’s approach is to serve customers who want a streamlined, optimized, lower cost version of DWDM gear?
CL: That's right. They want a DWDM product that gets their job done - most of them want a simpler solution because they have other problems to solve. We want them to be able to focus on their problems in their solution space and basically get out of their way. Often times you get a product with all kinds of bells and whistles and you're paying for these features whether you want them or not. With XKL, the customer is paying for what they want and what they need.
TR: What types of customers do you target?
CL: The solutions we have were initially designed and targeted to enterprises and operators that have router and switch nerds on staff. We introduced an optical layer product that fits seamlessly into that environment. From those customers' point of view, the simplicity is much appreciated and they like the ability to build and grow the network any way they see fit. Customers that have optical engineers on staff appreciate our product as well from the standpoint that we make it easier for them to focus on problems with higher data rates, more bandwidth demands, and dealing with fiber that has been in the ground for over a dozen years or more. While they will look at our product initially and say it is too simple, once they've deployed it, commissioned it, and are familiar with how it all works, they often realize much of that complexity is not something they need to take on. We enable that level of customer as well.
TR: In the market, what level are those? Regional providers?
CL: We started off with enterprise customer as our target market. As fiber has become more available and as the bandwidth requirements have gone up, our market space has expanded considerably, growing to include regional providers, internet exchanges, clouds, and data centers. We really don't try to get into the Tier 1s, which is a very competitive market that has entrenched equipment providers, but we find ourselves in this space as well.
TR: OTT video and the cloud have been reshaping the networking industry, have they affected the way you design transport gear? Or has it just mostly changed the type of customer that uses it?
CL: It's mostly changing the customers. At the optical layer, technology advances occur at a much slower pace than at layer 2 and 3. Back in the year 2000, 10G transceivers were just starting out and the promise was great, but it took 5-6 years before those devices were commoditized and the technology was to the point where they were readily available. Along came tunable TOSAs that were reliable and able to fit in XFP and SFP+ form factors, and this was a real technological advancement. 100G DWDM based on more advanced modulation techniques like DP-QPSK, has been similarly out there since 2008-2009 but it is available in large quantities at competitive pricing only now in 2016. With virtualization, you can take advantage of virtual machines to spread the compute power around. But that hasn't had the kind of impact at layer 1 that it has at layer 2/3. Disaster recovery activities have had a more significant effect because capacities continue to grow as organizations try to get their corporate goodies and intellectual property in a secure place in case of a major disaster.
TR: How do data centers and internet exchanges use your gear?
CL: There are two sides to that. One is that in a single location data center buildout you have your ingress and egress, and that has always been the traditional application for DWDM optical transport gear. But now we are starting to see companies planning ahead and building multiple data centers within a small region, sometimes right next door sometimes across the street, wherever real estate is available and inexpensive. They will provision a significant amount of fiber between those facilities, but even with that they are constantly running out of capacity. It’s not just for raw bandwidth needs, but also there's a complexity in management, organizations don't know what's connected to what. If they have a data center connected to one next door with an 800-strand fiber bundle, they have a complexity of trying to keep track of where those fiber ends are connected on each side. Optical transport gear can provide an active patch panel for them. But in order for that to be realizable, the equipment has to be affordable. We have seen requests and provided equipment for this activity, even for very short reach. You would think they would just put in more fiber, but that has diminishing returns and sometimes you don't have the ability to do it quickly. The other side is in PoPs and carrier hotels with cross connects. We're finding organizations that are paying for cross connects, and as they add more and more fiber pairs, their monthly recurring cost is significant. Even though they're going just from one floor to another, there is a break-even point where they can decide to use DWDM equipment to minimize their MRC.
TR: Is that something you’ve been seeing a lot of in the marketplace already? Or is it future demand?
CL: We've been seeing more and more of that. Organizations are looking at how much they are paying for cross connects per month, and in a one to three year cycle, what's the cost to buy DWDM equipment and minimize the number of these cross connects. The costs are starting to equal out, because the DWDM equipment pricing continues to come down while the cross connect costs are staying flat.
TR: Has 100G technology reached the point of commoditization yet?
CL: The price is there right now. In early 2016 we’ll reach the point where a 100G coherent single lambda solution is equivalent to 10x10G. The real push, though, is the cost of the 100G router ports, which are still very expensive. So what they are trying to do is take advantage of the 10G and 40G interfaces on the short reach side but utilize the spectral efficiencies on the 100G line side. You're seeing quite a few products now, and more are coming, that allow customers to aggregate their 10G and 40Gs.
TR: How soon do you think we’ll be seeing 200G, 400G and even 1T optical gear deployed in production networks?
CL: You have to be careful when talking about that. When people referenced 100G when it was first coming out, there was little distinction on whether it was a single wavelength or 10 wavelength 100G. One implementation for 200G is implemented using a single wavelength, 16QAM modulation scheme. Another uses two wavelengths at 100G each. As more advanced modulation techniques start becoming realizable in the optical domain, which is a difficult problem to solve, we’ll see more spectrally efficient solutions. For 400G, two wavelengths running 200G 16QAM on each wavelength, or four wavelengths running 100G DP-QPSK are available today in production networks, but the cost per bit isn’t there yet. Commodity 400G in 2017 is likely. Most customers don't want to pay for spectral efficiency at today’s prices to get a 400G solution. 1T involves more integration, reducing the power consumption needed for multiple wavelength, advanced modulation approaches. This will likely take until 2020 or later before this level of integration occurs.
TR: What products is XKL working on right now?
CL: The 100G DWDM solution that XKL has, but hasn't announced yet, is a coherent single wavelength solution with a secret sauce that is not available to the industry at layer 1. It's basically what can you do with Ethernet to try to make your optical network more efficient and make more effective use of those wavelengths. You're seeing a real proliferation of Ethernet. Data rates used to change by a factor of 10. Then along came 100GE and people had a hard time dealing with it for longer haul transport, which opened people's eyes to the fact that this isn't going to be quite as easy to get data rates up by a factor of 10 each time. So an attempt was made to implement a direct-detect scheme at 40G. This turned out disastrous. So in addition to 1GE, 10GE, 100GE, 400GE, the IEEE is about to finalize 25GE and 50GE standards. In all of these different Ethernet standards, if you look under the hood, Ethernet is Ethernet. It has a header, VPN fields, a MAC address, etc. Each standard is running at a different clock rate, however. Our solution will take advantage of that.
TR: What new capabilities do you think customers are looking for today?
CL: The evolving market is not just about more bandwidth and spectral efficiency, it's about space and power issues. The data centers are clamoring for more green equipment, devices that won't blow their power budgets. Racks are half full because they've run out of power and therefore cannot install more equipment. If you look at some optical transport solutions out there now, a popular equipment vendor has a solution that includes a controller card consuming 120W. You're not even transporting any customer data yet and you're at 120W. You start looking at the fan trays as well, it's a huge power budget. Customers are asking to get the job done with a much smaller physical and environment footprint. You're going to see a lot more innovation in this area.
TR: What has been driving increased power requirements and what can be done to slow it down?
CL: The laser technologies themselves have been advancing, but at a pretty slow pace. That is part of the power budget. The problem arises when you start throwing in every feature under the kitchen sink. Equipment vendors have to sit back and understand what the requirements really are for optical networking. Do you really need all the technologies you are throwing in there? Another issue is that some of these problems are hard to solve. A 16QAM solution requires some signal processing, and those chips are upwards of 40W per 100G. We have to be more careful when engineering at the component level and system integration level. That has to happen. You can't just get the functionality down and not worry about what the costs are. You have to think ahead.
TR: Thank you for talking with Telecom Ramblings!