New technologies always take longer to revolutionize an industry than the initial hype would suggest, and SDN and NFV are no different. But they aren’t going away either, and when carriers look to reengineer their infrastructure to take advantage of new technologies like this, they often bring in outside expertise like Datavision to help kick things off. With us today to give his perspective on the shift toward software-based networking is Datavision COO Mark Abolafia.
TR: How did Datavision get started, and how did you get involved?
MA: Datavision is a network integration and consultancy that was started by Mike McKeever in 2004. He and I actually worked together during the telecom boom/bust prior to that and had kept in touch throughout the years. Four years ago, we ran into each other at a Christmas party, and he was describing some work that he was doing at AT&T with SDN. He suggested I run it for him, and after considering it for about 3 seconds, said yes! So for the last four years, I've been building up the company’s SDN capabilities and practice across a number of different areas of competency and assembling a partner ecosystem.
TR: What steps did you take to break into that world and grow the business?
MA: We made the investment to join the MEF, and we hired a few very smart individuals who could help us get up to speed on a lot of the models-based approach and open source aspects of what SDN was becoming. And we also focused on winning business at places like PCCW, AT&T, T-Mobile, and Cisco.
TR: Are you also targeting smaller carriers, vendors, and enterprises?
MA: Yes. The large carriers are where the impetus for large investments are coming from - PoC’s are conducted, and production systems are deployed. But what we've also done is to address the Tier 2 through Tier N marketplace and put some boundaries around specific products and integrations to keep the cost and complexity down. If you do that, you can get someone into automation as a smaller business for significantly less than the big guys have paid for a full blown automated orchestrate- everything-on-the-planet type of thing. With that, plus our SD-WAN push, we're looking to provide the benefits of orchestration into the enterprise space as well. We've done some enterprise work but it's not currently our primary revenue stream. But given the level of competition and the extremely lengthy sales cycle that carriers have, we want to diversify the business. We’ve got a few things in the pipeline, such as SD-WAN and with Cisco NSO that will be bridges into the enterprise space. We will leverage our service provider experience and some pre-existing technology and development work to help productize it and deploy it more cheaply and quickly than some of the larger Tier 1 carrier work that we've done.
TR: What have you been doing with the MEF?
MA: The MEF 3.0 LSO framework has different interfaces that are named after various opera terms such as Legato, Allegra, Cantata and Interlude that represent interface points between different layers and levels and silos of orchestration whether it's within a single carrier or intercarrier communication. We were involved in a project that essentially took Carrier Ethernet 2.0 standards that were already baked into MEF and converted that to YANG models that could be standardized and used across different orchestration platforms such as TAIL-F/Cisco NSO and OpenDaylight. Both of those are orchestration platforms that consume YANG service models to abstract the services layer.
TR: What are you seeing in the enterprise space with NFV right now? How are providers and their customers approaching the space? How much of it is leveraging expertise like that of Datavision?
MA: What I have seen are carriers of different sizes and flavors offering a form of NFV to enterprises via some form of whitebox. At this point in its evolution, that carrier usually either had a company like us in there to get them started, but they are now hiring a ton of software developers and network engineers who are knowledgable in this technology. There's still that extra edge of expertise that we have, having done this across multiple customers, and that is what enables us to differentiate ourselves.
TR: How easy is it to find people with the qualifications to work on this stuff?
MA: Three or four years ago, it was quite difficult. But we had a few good people as a core and we were able to conduct train-the-trainer types of activities to get people up to speed. Three years ago, you couldn't find a single person who could spell OpenStack, much less get something to work. Now it is much easier, and I can find plenty of people to do OpenStack today. However, there are still some technologies like Cloudify and ONAP where the expertise is still very narrow. I wouldn't call any of it a commodity skill though, because it's a set of competencies that are difficult to learn and do right. One can be a mechanic about it and just say, "Yes, I have downloaded OpenStack and got it up and running and, yep, there it is." But there are other guys who are more like artists with it, and they are still rare (and we have them!).
TR: Where are new people like that coming from? Do you mostly get software people moving into networking, or from networking into software?
MA: Both. There are network engineers who, in the last couple of years, have said to themselves, "I better learn this or I'm going to be out of work." And there are software guys who are saying, "Hey, this is a cool new way to exercise software and that part of my brain. Let me learn some networking." So there's really been a lot of cross-pollination. But much of the newer talent we're seeing is primarily from the developer side. It's more difficult for a network engineer to become a good software developer. If you have a real good software engineer who knows not only the developing side but the networking side, then the code that they create, the solutions that they create are usually much better-engineered. So that's what I tend to look for. I would rather have a software person who is knowledgeable in networking than a networking person who is just about to learn coding – but both add value to the process that either could not on their own.
TR: How far along are we on the adoption curve for SDN and NFV and related automation technologies in networking?
MA: I think it's still at the beginning of the knee of the curve. This technology has plenty of room to run, because you're going to see every network reengineered to perform all this advanced functionality. The OPEX and CAPEX savings in the long run are just too compelling to ignore and the flexibility is too powerful.
TR: So what do you think is holding back adoption at this point? What needs to change?
MA: Well, there is always internal organizational inertia. For many, the existing network works well enough, so they’ll say that they’re going to migrate it, that they’re not going to do a whole forklift upgrade of everything, that they need to have the business drive it, etc. It’s not a case of “not invented here” syndrome, it’s more of, “we have to be sure that we understand this technology before we deploy it” and that's a process that take time. But we need to have application development itself change, because even if you have a software-defined network then why would you change anything unless you're going to change the applications such that they leverage the network to their need. And even then, there are a lot of different aspects that need to be engineered and thought through before you can wholesale move an enterprise over to it. I think that that's really the gating function.
TR: What technologies are coming down the pike that interest you the most? Are there other opportunities in there at the intersection of software and networking? Edge computing? Blockchain?
MA: In the near term, 5G and IoT suggest that edge computing, and analytics, could be big. I think you'll see us get further into the edge compute side because today that starts with something like a lightweight version of OpenStack. We've got a partner called CPLANE.ai that has a really interesting solution for managing multiple mobile edge computing locations, all through a single API. From that perspective, if someone walked up to Datavision today and said, "Hey, can you help us with this edge computing thing?" My first answer would be yes. The second question, from me back to them, would be, "How many do you need and where do you need it? As for blockchain, it’s early days for a non-cryptocurrency application of that. There are plenty of folks who are working on blockchain for smart contracts and things of that nature. Do I see it applied it to networking, with respect to either authentication or other security methods or payment methods? Sure. Is that something we're involved with today? Not at the moment – though we are always looking out at the next “big thing” in networking, we need to balance our investments with our capital resources that will show a more near-term payoff.
TR: So what hurdles do you see ahead in terms of what you want to do?
MA: Finding the people fast enough to keep up with the demand, which is strong, and being able to find new opportunities fast enough, versus the competition. When we started the SDN world, we were one of maybe two or three organizations in the country who could do all this stuff. Obviously, with the maturing of the technology a little bit and obviously a lot more people getting trained on it, we're certainly not the only game in town anymore. We like to think we're a cut above a lot of our competitors in the SDN space due to our five-year-plus experience in operationalizing it for tier-one carriers. And for a small company like Datavision, with only 130 plus or minus people, we look, smell, and act like a much bigger organization with a lot more expertise than one would think – with the added bonus of being quite agile in responding to our clients. We’ve got some really good people who have been with us for a long time. We hope to continue that strong track record.
TR: Thank you for talking with Telecom Ramblings!