I was at the ODL Summit in Seattle September 11 and 12 and got to listen to the keynote from Microsoft IT and Project Atrium on UC SDN (the keynote session info can be found here and the video recording of the keynote can be found here). (Click here to view the detailed video of Project Atrium mentioned in the keynote.)
The keynote started with Brent Hermanson, who leads the Network Infrastructure Services group for Microsoft IT. Then Gert Vanderstraeten, who is the Network Architect in Microsoft's IT went into their technical initiatives and how it relates to UC SDN. Next followed was Dr. Bithika Khargharia, Director of Product and Community Management at the Open Networking Foundation (ONF), who leads the charge with Project Atrium. The ending of the keynote was very fascinating as they see the need for UCaaS clouds communicating to service provider and enterprise networks as being worked on by the IMTC.
Right after the session I had the opportunity to interview the Microsoft IT and Project Atrium team with a special attendance from ODL Executive Director Neela Jacques. Also in the interview is Zac Smith, Director of Architecture Microsoft IT, and Denis deRuijter, Chief Developer for Project Atrium. Hope you enjoy this very insightful interview on the challenging environments of IT and real-time media applications.
Pascal Menezes: Hi, this is Pascal Menezes. I am here at the OpenDaylight Summit in sunny Seattle. Who just came off the stage from the keynote is Microsoft IT and the Project Atrium team. I am really delighted to also have Neela Jacques, the Executive Director from Open Daylight. Neela, what did you think about the whole keynote from Microsoft IT and the Project Atrium team?
Neela Jacques: Sure, absolutely...My name is Neela Jacques. I am the Executive Director of the OpenDaylight Project. I gave a keynote to launch this event yesterday called, "OpenDaylight Has Just Got Real." It has really been the theme of the entire conference. It was great to hear from so many adoptions of SDN; whether it is service providers like AT&T, or major web scale companies like Tencent, all talking about how they are leveraging the ODL SDN platform to really solve real business problems.
Pascal Menezes: Great; awesome; so what did you think of the Project Atrium and Microsoft IT's presentation? What did you think as the executive director?
Neela Jacques: I was blown away. One of the things when we started the OpenDaylight project that everybody kept asking about is, what are the use cases? What problem are you actually trying to solve? While I think we all believe in SDN; it was a little bit of the field of dreams. If you build it, they will come. Build a broad platform for SDN, and there are all of these problems that people will start solving on top of it.
I think what I heard from Microsoft IT was a problem that I do not know anybody would have articulated two or three years ago. When I first talked to Brent who gave the keynote; and he shared with me this vision of, "hey, I want to save money and serve my mobile users better."
Okay. That sounds like everybody else. But then he said how; which is I want to destroy the corporate network. I just want to go Internet first. It made so much sense. But really, the key thing is that they can actually do it. This is not a pipe dream. This is not something in five years. They can do it now and the technology is there. We have groups like the ONF Atrium Enterprise project that have actually taken open source software and able to deliver on this vision. I think this really what is so exciting about the keynote.
Pascal Menezes: Yeah, I totally agree with you. One of the things is, in Open Network Summit in March or April; I cannot remember. Dave Ward, CTO of Cisco, talked about the idea of the Reactive Networking where applications program the network. I have seen this a couple of times now. People are starting to talk about applications working with networks and communicating in a machine-to-machine way.
What do you think on what the Project Atrium presented? Do you think this is kind of matching that vision? Are we in a whole new kind of model about what SDN will do in the way of actually working with end user applications versus orchestration applications?
Neela Jacques: Yeah. In many ways, what we have done here with the OpenDaylight Project is the ability to create a level of abstraction that was not there before. An ability to have greater management and control over every part of a given network.
That is what comes to my mind when you hear Comcast yesterday talking about, "hey...we can dynamically provide Comcast Digital Voice. Or, Microsoft IT saying we can optimize Skype for Business. I think both of those use cases are about abstracting a way the network which then ends up solving a very specific problem for end users out there; which really makes all of our ability to communicate with each other so much better.
Pascal Menezes: Yeah. I totally agree. I totally agree. Zac, welcome.
Zac Smith: Thank you.
Pascal Menezes: I know Brent could not make it.
Zac Smith: Yes.
Pascal Menezes: He did the keynote but he had to leave. I know you worked closely with Brent.
Zac Smith: I do.
Pascal Menezes: Zac, from Microsoft IT - why don't you tell us a little bit about yourself? Then, maybe I have a couple of questions; if you don't mind?
Zac Smith: I am Zac Smith. I am the Director for Microsoft's IT network architecture. We deal with the strategy for our global corporate networks but not the commercial side like Azure, Hotmail, and Bing as they are separate organizations. But, for the 700 and some odd sites and multiple data centers around the world, we are the organization that manages and engineers the network. Then my organization specifically drives a lot of these crazy strategy ideas that we have that we have been talking about today.
Pascal Menezes: Well, that's great. You guys are certainly massive. I mean, Microsoft has got a massive footprint. It is great to have you really stand up and talk about how Microsoft IT sees the use of SDN. Some of your key initiatives that Brent talked about on reducing costs and getting to automation are interesting.
How do you see this fits into making the network better? How do you think that this will be achieved? You talked about your initiatives on using the Internet first and all of these other things. If you can just talk a little bit more about how Project Atrium fits into this world that you see? How does SDN also specifically fit into this?
Zac Smith: Okay. Somewhat more generically, I mean, we are talking about enabling applications to communicate with the network to communicate intent, right, and have the network respond. In reality, we have been doing that ever since we have had networks. It is just that the programming interface was people and details.
Pascal Menezes: Yeah. It is exactly.
Zac Smith: That has always been there. But, there is a natural conflict that occurs between the strategies that we want to drive internally within our networking organization; and the things that we need to do to enable our business partners. We are constantly having these discussions about where do we apply our resources to help a product group or a business unit go deploy a call center in a building that was not set up for it? Or, is it to actually go and move our core routers and set it up for label-switched routers instead of a services router?
One of the things that we have talked about with automation being an accelerant for our organization; I think that is pretty clear to people. But I actually believe that enabling applications to communicate intent to the network, and have the network respond automatically, is also a huge accelerant for an organization like ours. Because we can get out of being the programming interface for the network and enable groups to do it for themselves.
Then we get to go apply our resources to make the network just function better. Really, just focus on how do we move bits from point A to point B around the planet? That is one of the things I am really excited about long-term for our network is enabling more and more of those scenarios. What we are talking about today with Project Atrium and specifically the Skype SDN interface is really enabling us to support key scenarios which allows our network to support collaboration, and our voice platform much more efficiently.
We talked about moving to Internet first. The question is how does this support that? That is, we would love to just give users an Internet connection in an office. Say there you go; use remote access to get back to your corporate resources but everything else, all of your workloads are in the cloud, so march on.
But the reality is with all of the various connectivity capabilities out there in the globe; as well where certain product and features are. We cannot do this and maintain a high quality experience. Project Atrium enables us to do this in a really smart way which enables us to change our network topology but still maintains a high quality of service for our UC platform, which is obviously Skype for Business.
Pascal Menezes: Right. The thing is I think a lot of listeners like our readers might say, "Well, that's Microsoft IT, a massive org. Do you think the same issues apply to the smaller businesses? Do you think that the pain point is even worse for these smaller businesses where they are trying to deploy voice, video, and make it run right? They don't want to be in that model of supporting that. They just plug in, go, and have a high user experience; and not have to worry about it. Do you think that these smaller businesses have a higher pain point?
Zac Smith: Yeah. I mean, I would say that they probably have similar pain points. While we have massive scale, and we have a fairly large organization; we still have a limited number of resources to go and solve these problems. The demand signal for us exceeds our capability to deliver at any given time.
I would assume that is the same for almost any networking organization out there. Getting the networking organization out of the way of the application, and being able to configure the network is going to be beneficial for whomever is doing it. In effect, outsourcing that configuration work to something like an SDN controller or the SDN API for Skype for Business is great. It is beneficial for any size organization.
Pascal Menezes: Okay. Thank you Zac, very insightful stuff. Now, we have the Project Atrium team. Thanks guys. What did you think? As you embark on Project Enterprise, why did you go there? Why did you take this on? It is kind of interesting to also learn about why Project Atrium even started and where you saw the opportunities.
Bithika Khargharia: Yeah. Pascal, thank you for this question. It is kind of an interesting journey with Atrium. We really kicked off this work with ONF coming and saying that - how do we really make SDN real? Which means, how do we really take it more out of science and make it real world abundance?
That is how we started with Atrium. We built an Atrium platform. That was the first thing. We did some interesting things with that regarding a few trials. Then, as I started scouting the landscape, it seemed like there was a real need and an opportunity to take more SDN deployments into the enterprises.
This use case came as a great sort of springboard to bring open source code for companies like Microsoft IT and potentially other enterprises. I think that is how really the whole journey started.
Pascal Menezes: Yeah. I mean, I think this use case is just a really use case for SDN in general where you have end user applications like this talk to networks. I want to ask you a question, though. How do you see moving forward working with ONF and IMTC? How do you see that maturing? Everything is going to the clouds as we are seeing Office 365 having Skype for Business. We are seeing many cloud offerings that are UC-as-a-service. How do you see this fit into Project Atrium?
Bithika Khargharia: Well, I mean, I think there is a 100 percent alignment into that. With the Skype for Business use case, we are considering an on-prem scenario with enabling the Skype for Business on-prem as a first deliverable. But we, probably in the future, we are also considering a scenario where something like Skype for Business will be sitting in a cloud. We want to be able to provide the Skype for Business connectivity through one or more service provider networks in between, connecting right up to the enterprise.
At each instance of this network domain - so, at the enterprise, we would have Atrium Enterprise running. That is the stack that we described and just talked about. Then at the service provider or domain, we would have an Atrium Service Provider version running. With Skype for Business in the cloud, we will be able to communicate requirements into each of these domains in a completely controller agnostic way using intent NBIs.
We will be able to provide the enterprise the same quality of experience, and diagnostics end to end. I think the cloud use case sort of naturally fits into this. That is where we are actually taking advantage of the work that the MEF is doing with their Lifecycle Service Orchestration project; so, writing to those APIs to make sure that the Skype for Business API just work through standard LSO APIs would provide that quality of experience end to end.
Pascal Menezes: Alright, and with IMTC?
Bithika Khargharia: With IMTC, it is basically the use cases that we picked up for Project Atrium Enterprise. They are pretty much from the IMTC specifications on Quality of Experiences and Automatic Diagnostic services. Then, the third one is essentially the cloud piece. We are just looking to this using the new IMTC cloud use case specification coming out.
Pascal Menezes: I got it. What is your view of intent NBIs? Can you explain what that means a little bit better? Because is this UC SDN interface an example of an intent NBI (Northboound Interface)?
Denis deRuijter: Yes. Let me say a few words about the intent from a developer's perspective.
Pascal Menezes: Great, Denis.
Denis deRuijter: Denis deRuijter, yes, I have been working as the lead architect for project Atrium Enterprise. I have been working on computers and networking for decades and I have seen the advances in technology. But it has always frustrated me that we cannot always innovate as fast as we should. The reason is, if we look at networking today, scale and flexibility is needed. But they are not friends, right. One is in the way of the other.
The way to fix this is by intent. How can one express in very simple terms what they want, and let the logic underneath take care of it? You don't get overwhelmed by these tremendously large numbers. Then the analogy, I would say. If you fly a small Cessna today you would operate the pedals manually by hand. That is all okay. But when you fly in a large airliner you need a joystick. The joystick talks to the computer. It handles the flaps, etc, so you have a level and direction.
Now, we have this same analogy growing in the SDN infrastructure. But what is missing is something on top that makes it very easy to steer. We are lucky that while intent is academically very interesting, we now have the opportunity to tie that to an actual business use case to show what this can do. How it can make a difference.
Pascal Menezes: It makes sense. It makes sense. Gert, you were on the keynote. I really liked your piece on the SDN front. A lot of people are talking about SDN for service providers which is very clear. Service providers are all transforming their networks to be more cloud-like in which SDN NFV and orchestration are playing a role in making this happen.
I think on a service provider front that is very clear. But on the enterprise, is there a retooling that is going to really happen? Over the next one or two years, are we really going to see this retooling happen all over the world? Will the enterprise take on this transformational technology?
Do you think this is actually going to start showing up in the enterprises? Do you have any thoughts?
Gert Vanderstraeten: Yeah, Pascal, so obviously, there is a whole lot that goes with SDN or NFV. It is not going to be something that we can do overnight. I think; because we run a network every day, we have to meet our quality requirements towards our end users. We will probably see SDN coming in step by step.
If you look at the specific use case, we talk about today; I think it is more of a baby step towards a real SDN solution. There is a first step. It is a full step. But it is something that we can very easily take and use in a brownfield deployment. We do not have to go out into an office and rip out the infrastructure.
Pascal Menezes: Yes, right.
Gert Vanderstraeten: We can add this little small piece here. We can show the value. This will give us the necessary experience to move forward and take the next step. At the same time, we can start showing some cost savings and getting more efficiency gains.
Pascal Menezes: I got it. I got it. It makes a lot of sense. Why OpenDaylight? Why do you believe in OpenDaylight versus - a lot of proprietary SDN controllers including Cisco's APIC-EM. Why do you think the ODL makes a lot of sense in an enterprise as an open kind of environment? Why is this important for you as Microsoft?
Gert Vanderstraeten: Right. We do have a number of different vendors into a network. We obviously, like every large enterprise, have the big Cisco shop. But we do have other vendors. We have HP. We have Arista. We have Fortinet. We have Palo Alto. We have F5. We have a whole range of different vendors.
We do not really feel that we want to be locked down into a proprietary corner going forward. The goal is really to use an open platform that we can use to match more than just the Cisco environments. It does not mean we are not going to look at an ODL commercial release. We will probably be looking at an ODL commercial release partnering with the vendor that has a solution based on OpnDaylight.
Pascal Menezes: I got it.
Gert Vanderstraeten: Like I mentioned in the keynote. Some of the vendors might have specific modules already written on top of it. Then it does not make much sense for us to reinvent the wheel. In fact, we are Microsoft. We do have a lot of good developers in-house. But, still that does not make us more efficient or more effective to kind of reinvent the wheel.
Pascal Menezes: I got it.
Gert Vanderstraeten: It still allows us to write our own modules and add them on top of the existing solution again.
Pascal Menezes: Well, I have got to say your use cases on the SD-WAN or wireless first makes a lot of sense. I think that this can drive so many opportunities in using SDN. I love it. I want to thank you guys. Again, brilliant, and I really loved your keynote. Thank you very much again Neela, for a great conference, as this is just an incredible amount of information from the show. It just completely blew my mind of how much OpenDaylight is being used in so many areas; and not just enterprise, service providers and everything. Hats off to you for executing.
Neela Jacques: Thanks. I wanted to share. Really, it is about the community. What I am most proud of are the hundreds and hundreds of people downstairs. They come and say they had a great event. I asked them why. Is it the content or the keynotes?
Yes somewhat, but it is also the connections they are making with each other. People feel like they are part of something that is bigger than just their company just making money on a specific product. I think this is our greatest achievement. That really gives me hope despite all of the challenges. That we are going to continue to tackle bigger and bigger problems.
Pascal Menezes: Yes, I totally agree. SDN is like a child. It is just trying to grow up. I think we are at that inflection point where the child is kind of growing into a teenager and now going into a young adult.
Neela Jacques: Absolutely -
Pascal Menezes: Thank you all for this great session. Thank you.