Open Daylight Control of Converged Packet Optical Can Enhance Network Efficiency and Revenues for Service Providers
By Mike Capuano
VP, Corporate Marketing
Infinera is a member of the OpenDaylight Project, a community-led and industry-supported open source platform to advance Network Function Virtualization (NFV), and Software Defined Networking (SDN). At Insight Infinera 2014 on September 18th, we demonstrated control of PXM, our new packet switching module, in the DTN-X from an SDN controller running the OpenDaylight Project’s open source software release.
I believe that the combination of NFV, SDN and efficient packet optical transport will provide service providers the most efficient network architecture to deliver and manage network services. NFV allows previously appliance-based network functions such as intrusion detection systems (IDS), session border control (SBC), broadband remote access server (BRAS) and many others to run on top of virtual machines on commodity-off-the-shelf x86 servers (COTS). SDN separates the data plane from the control plane, thereby enabling the underlying infrastructure to be abstracted and directly programmable with the potential for new revenue-generating applications and more efficient network architectures.
Vice President, Hardware Development
Like many stories, this one starts in a bar. In a bar, in January, on a Monday night. Unlike many bad stories, this night wasn’t particularly dark nor stormy. But the night was memorable, as it was a real beginning of what would become a new product family.
We were preparing to meet with a customer, a major internet content provider, to pitch an idea for a new product. The account manager suggested that we float the idea the night before, in a more relaxed social setting, when meeting the customer in a bar. It was here that when we told the customer what we were thinking, we first heard “no, that’s not really what we want. But if you could make this modification, then maybe we would be interested.”
The next day we spoke about “this modification” and there was clearly mutual interest. A month later we met again, with more details, and showed our development timeline. Over the next few weeks, we met with another internet content provider who also expressed interest in our proposal, but they too, wanted a modification. After some consideration, we determined that their change request was something we could accommodate without compromising the first customer’s requirements.
With strong feedback from two potential anchor customers, the product requirements were solid enough to begin the hardware design. We leveraged proven technologies, including our Photonic Integrated Circuit (PIC) and our AOLM2 super-channel line modules so that we could get a quick start.
We continued to meet with potential customers. Another internet content provider. A data center operator. Another data center operator. Sometimes our product concept matched the customer requirements, but a few times they didn’t fit perfectly. We had to prioritize, make small tweaks if possible, and most importantly we had to stick to our committed schedule. We knew that no single product can meet everyone’s requirements. But we were pretty confident now that we could really solve a number of customers’ problems with our idea.
Three months after starting the design we had a technology test board up and running in the lab. One month later, we had a mechanical mockup that conveyed all of the key physical features and really highlighted the compact size. We showed the mechanical mockup to several dozen companies at NANOG 61, a technical forum attended by many networking companies and data center operators. The feedback was resoundingly positive.
Soon, we had our first product prototype in the lab. Within a few weeks, data was flowing through the entire system and we were able to close metro-distance optical links, carrying 480 Gb/s of traffic. Immediately we were able to show the potential customers the product in operation. Even at this stage we were still soliciting feedback for the final hardware design, and a key potential customer gave us another change request. This change would make the product easier to service in the field. While one part of you hopes that your designs don’t need to be tweaked, it was gratifying to hear this change request, as it showed that a potential customer was seriously thinking about how they would use the product and how they could avoid operational issues, even before we were ready to ship.
The working prototypes really expand the type and number of customer conversations you can have. Potential customers see that the product is real, and not just a collection of slides. And the number of positive customer conversations we had fueled the development team. They knew that they were working on something that was different, that was innovative, and that was really needed. The development pace picked up even more.
I won’t say that the story ends in a bar, because the story is just beginning, and I don’t want to convey the impression that we spend all of our time in bars. But we did have another meeting in a bar. In a bar, in December, on a Monday night, less than one year after our first meeting in a bar. This time, no customers were present, just the Infinera team that worked on Cloud Xpress. This time, we were celebrating the fact that a few hours earlier, we had just started shipping in volume, the Cloud Xpress.
After months of hard work, the entire company had come together and delivered from inception to general availability a whole new optical transport product that delivers unprecedented density, consumes relatively little power, and packs a powerful 500G of capacity in a tiny 2RU size. We had delivered the Cloud Xpress. In less than one year we had gone from product concept to shipping product, and I believe that the future looks bright.
For more information on Cloud Xpress and its benefits, watch as Infinera Senior Vice President of Cloud Network Strategy and Technology Stu Elby talks about unleashing cloud networks:
Proof Positive: Virtual Router Announcements Reinforce the Importance of Highly Scalable Optics and a Converged Intelligent Transport Layer
VP, Corporate Marketing
At Insight Infinera in September Infinera CEO Tom Fallon presented on two important trends that we are seeing in the marketplace that reinforce the importance of an Intelligent Transport Network. The recent virtual router announcements by Juniper and Alcatel-Lucent are proof positive that this shift is underway.
The first trend Tom discussed was the rapid adoption and growth of cloud services and infrastructure, enabling Network Operators worldwide to actively deploy and architect their next generation networks with both Network Function Virtualization (NFV) and Software Defined Networking (SDN) in mind. Network Operators are not only deploying their own dedicated data centers but are also turning central offices into data centers. Cable operators are doing the same with their head-ends and distribution hubs. Network operators have begun to take advantage of the massive deployment of commodity x86 processing power to virtualize network functions that were previously housed in dedicated appliances. Many service providers have stated that this network architecture could dramatically simplify, reduce cost, and increase the flexibility and scalability of their networks. Most thought that NFV would apply to Layer 4-7 functions such as Session Border Controllers (SBCs) and Evolved Packet Cores (EPCs) and Layer 3 only for low speed routing such as customer premise equipment (CPE). However recent performance tests and announcements by Brocade, Juniper and Alcatel-Lucent of virtualized edge router software-only implementations may portend the future of a “soft edge” network where edge routers are deployed on x86 COTS hardware.
The second trend he discussed is visible when you start to look at high speed and high capacity networking.
Infinera Product Evangelist
It’s been 10 years since the commercial introduction of a PIC-based transport network platform. Thinking about this, I combed through some archival video footage and came across an internal video that included interviews from some of the very first people to come on board at Infinera in 2001 with the dream of creating a large scale commercial Photonic Integrated Circuit (PIC).
Amongst those interviewed was Dr. Chuck Joyner, a former Bell Labs guru and keen bonsai grower. Some of you may be aware that back in 1969, the folks at Bell Labs were the first to come up with the idea of a PIC, although they didn’t really capitalize on that innovation in terms of transforming it into a product.
Chuck recalled his initial discussion with Infinera co-founder Dave Welch on the prospects for developing a commercial PIC, and I’d like to share his comment with you word for word…
Systems Engineering Director
Hello from New Orleans. It’s cold down here. SCinet is up and running a cool 1.5TB of capacity.
The theme of this year’s Supercomputing conference is HPCMatters (High Performance Computing). The technology shown throughout the show floor is all leading edge. Innovations from the Supercomputing community have far reaching impact – to every corner of science, to investment banking, to the discovery of new drugs, to the precise prediction of the next major storm.
Last night I watched a demonstration by GÉANT in which an astrophysicist in Cambridge controlled a satellite array in Australia while collecting and processing real-time data at his terminal.
And it’s not just about geeky scientific things either. Supercomputing has produced a number of short videos showing how High Performance Computing is used by companies and research organizations to make better products and new discoveries. This video shows how Procter & Gamble use HPC to make better shampoo and diapers. And as I said, it’s cold here this week. But fortunately that helps to add new data points to climate modeling, possibly one of the most difficult problems in the world. This video also shows why climate modeling is important.
Hopefully they can tell me when it’s going to warm up again.
Systems Engineering Director
November 17 begins this year’s Supercomputing Conference for high performance computing, networking, storage and analysis. This will be the 27th annual conference, and the seventh for which I have been a contributor.
Supercomputing has always been a unique conference. It brings together a diverse set of people and companies, researchers, educators, students, and collaborators dedicated to furthering science.
My part of Supercomputing has been to provide some of the communications facilities linking the many booths on the convention center floor with research and commercial networks around the world. Over the last seven years, I have seen the SCinet network grow from about 200 Gigabits of live capacity using 20 x 10 Gigabit client services to this year’s record of over 1400 Gigabits using 14 x 100 Gigabit client services. And this is on systems that have the ability to support a total of over 30 terabits of data carrying capacity.
The SCinet WAN team has always tried to use the most leading edge systems, often in pre-release versions from the WAN vendors. The last time the Supercomputing conference was held in New Orleans in 2010, we built the first ever long haul 100G facility from New Orleans to Chicago using specially built 100G LAN modules. Amazingly, it only took five days from shipment of the modules to bring up that 100G circuit.