November 18, 2019

Driving Open Networks and SDN

In this episode of The Broadband Bunch, recorded at the Broadband World Forum 2019, we speak with Aseem Parikh, Open Networking Foundation, Vice President, Solutions and Partnerships. The Open Networking Foundation (ONF) is a non-profit operator led consortium driving transformation of network infrastructure and carrier business models.  Aseem talks about the state of software defined networking, explains how to operationalize standards and offers his view of the industry today


ETI Backend Solutions

 ONF: Broadband Operator-Lead Mission

Pete Pizzutillo:  There are a few perspectives on openness, standards and software defined networks (SDN). ONF is positioned as operator lead; can clarify that for us?

Aseem Parikh:   What operator lead means - the primary places where technology gets deployed. Which then customers or their consumers, their customers or consumers get to use, whether they are individuals or enterprises or government agencies. All of us, obviously, over the time have used more and more of Telecom services.  By being operator led, there are pluses and minuses frankly. Pluses are, they know exactly what their customer base wants and what the pain points are. A simple exampe, which is evident for anyone in this industry is the traffic growth and the endpoint growth from IOT or mobile cases has exponentially grown over the last 10 years. You can look at every vendor's or every Telco's chart to show that. But which is also obvious is that revenues have plateaued or decreased.  It doesn't take a genius to figure out that when your OpEx and CapEx, which are proportional to the growth of your data, are also going kind of symmetrically in that exponential manner and revenues are not increasing something has to give.  John Donna went off AT&T about three years ago when he just announced the domain 2.0 project, which AT&T was on about, said that if we don't do it in 10 years, we will be extinct.  10 or 20 years, whatever time frame he said, I don't remember. But that is exactly the problem we are trying to solve.

That's why having an operator led mission is important. The Cisco, Nokia, Ericssons have done business very successfully for the last 20, 30, 50 years. And there is no reason for them to want to give up those large margins, their way of business, putting everything in hardware, may get close proprietary. And every time you have a new protocol, they add stuff to it without taking it away.  What operators have told us, and this is again when they sit on our board, a typical operator uses less than 30% of a device's capabilities, but they pay for all hundred percent.  Because there is no way of picking and choosing what you want.  If you're an MPLS user, you use MPLS but everything else in this box is capable of doing anything else, right? Whether it's layer one, layer two, layer three, layer seven, but you pay for it.

The disaggregation of monolithic architecture is what came to be and that is the key. In Malcolm Goldsmith’s book, What Got You Here is Not Going to Get You There, the the subtitle of that was what made you successful is not going to make you more successful. Those mechanics now are not going to work.  For this industry, what has got us here, the monolithic way of building hardware, it's served very well for the industry for the last 30 years, but for the next 30 years, our claim is and our thesis is that that is not going to scale.

How did you end up in the role that you're in today?

Aseem Parikh:   My first startup was acquired by Novell, which is where NetWare, the LAN networking opting system came out of. And that got me into the networking space. I was charted to make Novell speak into TCP/IP was the architect of NetWare. I sat on multiple panels at NASA, and other places, which got me more thinking about the real impact of networking.

Then the software entrepreneurship bug struck me. I decided to look at the broader side of technology; not only the technology as an implementation, but the impact, and how I can move the needle. We started a company, which was acquired by Cisco which basically created the Cisco Transport Manager, which is a large network management platform for all their optical gear. This was when Cisco was buying optical companies every other month. I was on eight due diligence's, and we paid out of over $10 billion in the market to acquire companies like Sarandon, Monterey and Keaton, and Pirelli in Italy, and a bunch of these companies. Every time we acquired a company, there would be a plethora of hardware equipment which my platform had manage. By the time you finish integrating the software for managing one device, the device software would change, and then you would start all over again.   Soon after that, I said "Okay, maybe there's a better way of building systems."

My next startup was in silicon company and I didn’t know anything about silicon. But the founder of the company said "I have enough silicon people. I need to understand what the application of the silicon is going to be." I became a product manager and then the general manager of that business unit. And we built some interesting silicon to bridge optical and IP domains together.

Then I did another startup in the big data space and looked at Telco data, operational, as well as for the business purposes of ensuring that we can enable a lot more agility in decision-making, even in the Telco. The Telco industry, as all of us know, has pretty long-term view. Things are difficult to change. They have been doing things one way for hundreds of years, or tens of years in many cases. And in that scenario, it became clear that any help they can get with making decisions faster would be useful. We pitched this to Sprint, Verizon, AT&T.  That was acquired by another company out of France called Thales.

And that made take a pause and say, "Okay, I've been doing the startup thing and been on that treadmill for a while." I wanted to look at things more holistically. And I had met Guru Parulkar, the executive director of ONF today when we are both at Cisco back in the late '90's, early 2000's. He had been acquired, and I had been acquired, so we worked together. And then I had followed Guru's trajectory, and he went to Stanford, that's where SDN started as a clean slate project. And we said this is an interesting way of looking at the world. If you bring software, you bring agility anywhere you bring software.  It's easier to spend software than spend silicon or a hardware device. That was the genesis of kind of my looking at that world. And, at the same time, everyone has seen what Linux had done to open source from the operating system perspective and from the computer perspective. We wanted to marry those two.

And there were these three organizations, which came out of Stanford's The Clean Slate Program. Two of them, one was ONF, the old ONF, which was a standards organization. And that's where standards like OpenFlow were created. And the goal was just to evangelize SDN as an alternative way of doing a technology, networking. Networking in all cases, Enterprise as well as Telco. And then the second group which was created was called Open Networking Lab.  Open Networking Lab was the one that just focused on software and creating open source software and primarily focused on the Telco, very robust high scale, high availability environment, which a Telco requires. It's a very different environment than an Enterprise Software. Over four, five, six years, these organizations co-existed. One did evangelism and standards, one did open source and then the board, which is all led by operators felt that really our charter of evangelizing SDN is now already over.  Anyone doing networking today uses SDN - whether they really do SDN or not is debatable, but at least the theorists in Stanford and Berkeley and Princeton don't think most of these products are real SDN and that's a different discussion.  So the new ONF, which was created about two years ago is mostly focused on the open source part of their business.  We a team of 30 people which create architectures, reference designs and exemplar platforms.  And to make sure that they are not influenced by any vendor. That's our key differentiation from other open source bodies, which are typically dominated by vendors.  And that's where obviously the skillsets are, the engineering talent is, but at the end of the day, each of those people who are working on those bodies are employed by someone else as a day job and there are biases and there are influences, and all of that. We wanted an unbiased way of trying to get the right technology, the technology because which is right to do, not because it's convenient to do for a particular X, Y or Z.


ETI Customer Churn

 Broadband Standards

Aseem Parikh:   We are not a standards organization and that is the key decision being made by combining the organizations I talked about.  The way it's been done for the last 30 years is you do paper designs and argue about every clause and protocol and every knob in there for years and months at a time and then you come up with a standard and then implementation starts.  ONF is turning that process upside down. You implement things first. You write software first if it works and the industry sees that there is potential of getting this, then you start standardizing the interfaces in between those disaggregated software components. This is what we call software defined standards. and that's another interesting perspective of how you bring that to the market.

Pete Pizzutillo:  You mentioned your clients are typically tier one companies but a lot of our listeners are tier two and tier three.  Help us understand your strategy of focusing on the tier one there.

Aseem Parikh:   Developing software defined networking requires a change in multiple levels in an organization, like a Telecom organization. One is obviously going away from this monolithic hardware, I'm buying a hardware, I'm buying a box mentality.  Procurement teams are tuned to buy this version of the box and these speeds and feeds. The industry has trained itself on doing that well. When you start buying disaggregated components, by component, the pricing, how that works together is difficult. You need to fit into the current model and need someone to integrate those disaggregated pieces into a solution. So, and that solution then, if those pieces come from multiple places, another thing in the industry is there's this requirement by all operators to have one neck to choke when something goes wrong.  But now there are 1800 vendors they can call rather than 30 people. To fit into that model, we said the small medium who would have maybe easier a way of trying this new technology out in their networks technologically or from the proof of concept or a trial perspective, they would be very happy to do it. But when it came to deployment, they would have a lot harder time because they would have to transform this whole industry and they don't have the skills and the people required to do that within their organizations. You need large organizations like AT&T, Comcast, Verizon, and Deutsche Telekom to say, "Yes, we are willing to put at a hundred people team out of our thousands of engineers and smart people who are focused on driving this kernel of change in a very small location."

Another thing which I didn't mention earlier, maybe this is interesting for the listeners to understand, is our focuses the edge or the of the network and the excess part of the network and not the core of the network.  This is strategic for two reasons; one that's for every core there are lots of edge points. The problem of CapEx and OpEx gets multiplied.  When you think about the famous story about some reporter asking a bank robber, "Why did you rob the bank?" Answer is that's where the money is, right?  If you want to try to save CapEx and OpEx, you attack the problem there, where the problem is most acute. And that happens to be the edge of the network. The whole architecture of central office as a data center was an ONF invention. We wanted to bring the key benefits of what Google and Amazon of the world have done by creating large data centers in the cloud.

Using merchant Silicon, using device hardware, which is white boxed. They don't buy from the typical OEMs, their networking gear or their compute gear. They build their own through OEMs and others. We wanted to bring that same thing to the Telecoms.

Pete Pizzutillo:  I think the idea that you talked about the disaggregation is similar to microservices or service orient architecture in software development.  Where you decouple monolithic systems into serviceable components. Is that a good analogy?

Aseem Parikh:   They're, the microservices part is the implementation of that strategy. The strategy is first to decide that a router has this 2020 blocks it's made up of. And as I said earlier, in a typical Teleco, you use only 10 of those 20, or seven of the 20 sure.  I only want to instantiate in my network this seven out of the 20 functional blocks. Now, how do you implement those seven blocks, could be microservices, docker, VMs and so on. We obviously are a microservices cloud, like a cloud service-oriented architecture. But that's an implementation choice that's decoupled from the strategic choice of just disaggregation.

Pete Pizzutillo:  It's interesting, attacking the tier ones has two benefits.  One is that you see that there's an impact through the, a structural impact through this choice.  You're changing how people buy systems which in some companies is a hard to do.  You're also examining deployment options a tier one has some flexibility, some investment dollars and bodies to explore how that change really ripples through their business.  Whereas the tier two and tier three, it's much more of a bigger gamble because it's a larger percentage of the resources and transformational effort, if you will.

Aseem Parikh:   It is, and I think you rightly said, the bigger guys, once they have decided to go through this architecture and they start to think about production deployment, they need to now find the right supply chain. And the supply chain has to change. It cannot be the same old guys, right? What ONF does through its membership is creates a curated supply chain for every solution it brings to market.

And those suppliers look very different than a typical incumbent supplier for these operators. And that's another reason why the tier one operators who have lot more muscle in creating the right supply chains are more successful in at least being the first to adopt these technologies. And that's what our thesis was. And it seemed to have played out where Comcast has just announced in September that they have now gone live with the first fabric, which is SDN based in their network serving 20 plus geographies, their hundreds of thousands of customers with what we call Trellis, which is SDN fabric.

Pete Pizzutillo:  So you're out in front of the industry helping the buyers and the tier two and tier three by redefining and helping folks rethink that, whereas if somebody's in tier two or tier three had an innovative idea, it's an uphill battle for them to try to convince everybody up and down the food chain.

Aseem Parikh:   But at the same time, having said that, we would still invite those tier two, tier three individuals to come and work ONF. And ONF is an open source organization, it's an open organization. Any of their membership tiers at every level. You don't have to be the top tier member to be part of it. The seven operators were only the top tier members. But there are what, 30 others, 40 others around the world who are at the second tier. And every tier two can start playing with this, can put this in their labs and now that we are seeing the tier ones going to production with Trellis already with Comcast as I mentioned next year hopefully when we are back here in BBWF you would hear AT&T Talk Telecom and Deutsche Telekom having gone live with the SDN enabled broadband access.

Broadband Maturity

Aseem Parikh:   If I think about three years ago, if you had asked me this question, I would have said the risk is, I don't know, 50% chance that this would work, 50% chance it will not work. Like any research projects. Right? Or any new project. Today that block has gone to the other side where it is not a question of if it'll work. It is working, it is deployed, it requires change, it is hard. It is not easy for to go through the change. But if you don't go through the change, you'll be roadkill is another message.

Pete Pizzutillo:  When is that timeframe? Not to lead with fear, right. But there's a lot of transformational issues that if people aren't thinking about them today, they're already too late.

Aseem Parikh:   I think it is, especially in on both sides, on the operator side and the supplier side, right? The operator side, like we talked about the ARPU going down, especially in the tier two and that kind of economically upcoming regions, I'm thinking of people, places like China and Brazil and India, where a gigabit of mobile bandwidth costs peanuts and we in the US are used to paying $100 per month. They can barely pay two or three for the same or more because most of them consume lot more video on their mobile devices than they are at their homes. For those people and this other, the tier twos or tier three operators, if you don't change and you continue to buy from the same incumbents, at some point you're, it's the math is not going to work.

The supplier side the same way, right? If you stick to your original business of selling boxes and proprietary boxes and support services on top of that and all of that, if your customer is bleeding, they just cannot buy more.  Any project has two risks, technology risk as well as a market risk. And we are surely behind the technology risk.

I would be also honest and say the market business models have still not been crystallized of how an operator will work with the system integrator who takes this multiple open source projects together. And ONF is just one part of that. There's going to be things like ONAP and other projects from BDF and others, which people would have to put together to build the whole ecosystem of what is the operational systems and BSSs and OSSs and the excess gear and all of that. That would take a different, and that part is what you are to now wait and see in the next two years. When the answer is when it will be too late? By 2022 it will be too late if you are not already figured this out.  Because someone would have figured it out and kind of got it to working at scale. Right.

About the Open Networking Foundation

Aseem Parikh:   We are an open source organization and an open networking organization, everything we do is in the open. Go to our website, open networking.org there are links which lead to every project. Who is on our board, who is driving it. There are technical steering teams for every project which are open to all the members. And becoming a member for small organizations is a few thousand dollars a year to if you want to be on the steering committee of those things, and then it gets a little more expensive. But for a Telco spending half a million dollars a year when the transformational nature of that is in the hundreds of millions of dollars is what the bet is all these people are making. They're putting this money into a common pool. There's a common RND which is done through ONF in the ecosystem and the community, and the risks and rewards are shared.  Another key point to which you might want to leave the operators with is, all operators have 60-70% of their stack of technology is undifferentiated.  That's the only piece we are focused on. That your secret sauce, which is on the top 20, 30% you go and figure that out through your preferred partners.

We are not trying to open everything, this is open is not that we are trying to make the apple cart go into a socialistic kind of every, there's no money to be paid. Open is not free, open as pragmatic. And where it doesn't make sense where you're not differentiating yourself and if there's commonality we are creating open source standards and open source tools around it.