Webinar: Next-Gen Network Analytics and Automation with Ryan Lynn
Ryan Lynn, VP of Emerging Architecture at Trace-3 discusses the latest in network analytics such as streaming telemetry, machine learning, AI and how network practitioners can get a head-start to deploy them.
[00:00:00] Hello and welcome to this special edition of the network collective short take. Today’s short take episode is being sponsored by Anuta networks. Now this is the second time that Anuta has been on the show. And once again they bravely ask someone who uses their products to share their experiences with the Anuta ATOM platform. Listen in as we explore how multivendor analytics assurance and orchestration are being approached in the real world and how Anuta can help you achieve these goals.
[00:00:30] So Ryan before jumping into the topic at hand why don’t you tell us a little bit about yourself. Where are you currently working and how you came to be engaged with the Anuta platform?
[00:00:39] Ryan: That sounds great. Thanks Jordan. Thanks for having me. My name is Ryan Lyn, I started my career in networking over 20 years ago and I’ve spent about 13 of that 20 working on very large scale global networks and data centers. In the last seven years I’ve been with a company called Trice 3 which is a strategic technology partner that really focuses on emerging technology. I’m a VP of emerging architecture for our mountain states where I lead a team of consultants in really three areas: cloud environments which include private, public, hybrid multi-cloud which we all know that networking is a big component of those things. Next Generation security platforms. And then lastly data intelligence and analytics. And so my primary role is really to help clients solve problems by finding and using emerging technologies. And in the case of Anuta networks which we found a few years ago, we were really searching for technologies that could help clients transform their operations.
Challenges with Traditional Analytics Solutions
[00:01:39] Interviewer: Right. Well, before we dive into specifically you know all the great things about Anuta. Let’s just start conversations about products by framing the challenges that they’re looking to solve. I think oftentimes we’ll get the cart before the horse and sometimes we have solutions before we have real problems. And I don’t think that that’s the case here. So from your perspective what do you think are some of the challenges with the traditional models and approaches to network analytics and assurance?
[00:02:04] Ryan: Well Jordan I think if you look at where we’ve been and how networking has grown up over the years I mean we have had some early issues right. The most obvious one is the closed system of devices that we’ve seen from that vendor landscape. You know we have uniqueness and operations, capabilities the formatting of the data between the devices is different and what that’s led to is things like point analytics solutions where it’s very vendor specific. It’s been anchored to a specific format then and largely analytics in general and networking has been very piecemeal and very manual had to really kind of piece together different analytics sets across different types of solutions. The good thing is that things are opening up there and we’re seeing access to new data. But beyond that one challenge we still have a large amount of volume of data coming off the network.
[00:02:50] So the question becomes how do we extract that data? How do we ingest it into some modern platform? How do we draw correlations across different types of data where there’s you know kind of an inherent lack of standards? So once you have that data knowing which questions to ask and more importantly how to ask those questions is still challenging. We’ve also got resource challenges where we all know that we have limited operations staff trying to sift through mounds and mounds of data. And so really when you take a look at those three elements of people process and technology, trying to solve it right now could be a pretty expensive proposition.
About Innovations in Network Analytics
[00:03:27] Interviewer: And that’s really interesting because it’s just so true because there’s always; we have you know, it always feels like more data is the answer. I need this piece of data because last time I missed this piece of data and it wasn’t as you know that piece of data I needed to make a good decision. But the more data we have the harder it is to sift through that. So what do you think is changing? You know we said that these are the traditional challenges. We think that you know there are solutions out there that do better. You know what are some of the things that are happening now that says that this is a solvable problem?
[00:03:56] Ryan: Yeah I think the first thing to look at is in order to have effective analytics in any area specifically in network and certainly in the assurance area you have to have access to the data. That’s the first thing. So the first solution we’ve seen is that opening up other devices and being able to access the data through very common mechanisms things like API’s, integration points are so important.
[00:04:17] Right so we’re also seeing other mechanisms of obtaining the data that tend to kind of change the model that whole model we’re used to. And that’s things like streaming telemetry right. And so we’re seeing that rise. The other thing we’re seeing is just the prevalence of mature data platforms right now we have platforms that can ingest, transform, prep that data more efficiently and they’re not proprietary they’re definitely open. And then you know that allows us to do that analytics and insight and that’s made better today and easier through things like machine learning algorithms and data science and modeling. So I think we’re seeing a combination of several points coming together at the right time to really enable proper solutions in the area of network analytics and assurance.
About Streaming Telemetry:
[00:05:03] Interviewer: Well there are a couple things there that you know they’re pretty popular terms right now. But I think that they kind of lack some definition. So let’s take a couple of seconds when you say streaming telemetry what do you mean?
[00:05:15] Ryan: Well I think if you look at like how we get telemetry from the network, traditionally, we’ve seen this whole model right where we’ve extracted data from the device. I mean I think the easiest example to illustrate there is SNMP, which I’m fairly confident has been in use since the early 1900’s. Even more than that like if you look at modern API access. It’s also a poll mechanism right. So when you look at streaming telemetry it works more like a push model.
[00:05:42] Ok so the device itself is now initiating data exchange by streaming that data to the destination. OK. And we’re also seeing improvements in that data model structure. I mean it’s not to imply that SNMP didn’t have data structure because it certainly did. But we’re seeing a more modern approach to how that data is being modeled in formats like YANG and it’s really unifying network under a broader umbrella of technology because YANG is used all over the place inside the application and inside of you know computing and those types of things. So streaming, so like what’s the benefit of streaming? Right. So we should be able to get higher fidelity of the telemetry data in addition to a more real time view of what’s happening on the network which I think is really important for you know making decisions, especially if you’re looking to automate things.
About Machine Learning and AI:
[00:06:36] Interviewer: And that kind of leads us back to the point we were just making a minute ago that more data isn’t necessarily always better unless you can do something actionable with that. And I think that’s the other piece that you mentioned there was machine learning. That’s one of the other really popular terms right now I think that there’s a lot of definitions. I mean there are some things and it’s getting conflated with AI. I think the two are kind of interrelated but aren’t necessarily the same. What’s your take on that? What do you think of machine learning and AI?
[00:07:01] Ryan: Sure. So Jordan I’m sure if you turned on the TV right now or browsing any Web site like we’re going to see that A.I. is running the world. I think the term is a bit misleading right. So academically AI or artificial intelligence really is the umbrella term that describes a larger initiative where we’re trying to have machines mimic human attributes. So we’re looking at things like Vision, Robotics, language, learning. So you take the last one and that’s the area that most of the effort has been focused around lately, that idea of machine learning. So we’re not talking about AI as a very specific thing, machine learning is sort of a subcategory.
So if you look at well how does that work in context with networking. Well think about the way you and I learned everything we know about networks. You know it was from constant exposure to the network. We absorb absorbed data and then we use that to train ourselves to understand what those data points mean. Right. So more interaction leads to more data which leads to more learning. So machine learning in the same token relies on a massive amount of data to understand the patterns and really draw conclusions from what it’s seeing. So the data is sort of pumped through data models and again more data means more learning and once you train that model to understand that these are what the elements mean, then that algorithm is capable of either automating or with human help spotting patterns, making decisions looking for trends and things like that.
[00:08:31] Interviewer: I mean and my take on it is contextualizing that they are taking the data and were providing context around it because it’s no longer just a number it’s a number in relation to other numbers which is the way our mind works right. The idea is why I think we’re looking at AI. You know we think about our intelligence our intelligence being able to take disparate data points and somehow combine them and mash them together to make something useful out of them. That’s been the hard part right of how information and data in the computer world because it’s two unique pieces of data and how do you teach it to do that.
So I agree that’s kind of the stuff we’re heading there. I think AI is one of those terms that we’re getting there. The machine learning is definitely making a difference and we’re seeing it in more and more products. I think the challenge here right is these are really cool ideas as technologies. Like streaming telemetry and machine learning and AI we will get very excited about them. I think what’s been hard is how to make those things practical.
About Embracing the New Technologies
So as a practitioner or as an engineer as somebody who’s running a network today what does this mean like what does that practically mean like how do I both in the short term and in the long term. How do I take advantage of these things?
[00:09:43] Ryan: Sure. Well I mean I think you know it comes down to the maturity of the organization. I think that ultimately reveals the path that most of our clients end up choosing. You know take an extremely mature network operations group and couple them with very strong development, very strong data intelligence people and I think they’re going to take a more long-term approach. Right. And what does that mean? So it means they probably have a very highly programmatic network very open telemetry. They probably use automation as a foundation. They probably have extreme development maturity and able to write applications that query that data, correlate that data extracted. You probably have the right resources and have implemented the right processes and so clients who have these tend to go on an autonomous journey. They tend to want to do it themselves.
But I got to tell you Jordan like majority of the clients I talk to, they are just not at this level. Right. So to your point how did they get started? Well, they need fundamental help. They need fundamental help with things that we mentioned before, automation. They need help and just getting data analyzing it you know putting it on some platform and drawing conclusions from it. And they probably haven’t evolved their staff or their processes to that next level of maturity. And so in talking with clients, this is generally where they say well so Ryan, we’re really behind and we are probably not a good candidate to go out on our own with everything. So like how can we accelerate that time to value with network transformation and I think this is where we look to emerging platform players like Anuta networks.
About Anuta ATOM’s approach to Network Analytics
[00:11:18] Interviewer: So you have done some work with them. I mean so your role right as you explained at the beginning of the show. You’re integrators you’ve done this with multiple organizations. So based off that experience because you’ve, had your hands in it for a while. What is Anuta doing that helps customers take advantage of these better analytics and assurance and orchestration and things like this?
[00:11:36] Ryan: Well I think if you look at the approach the industry has had historically it’s been buy this tool for this, buy that tool for that, very much point solutions. And so clients like that type of value point solutions but they don’t like the lock-in, they don’t like the isolation. They don’t like the inflexibility. And so this is where Anuta ATOM really starts to show its strengths. Right. It’s a platform approach which you know allows for a very solid foundation of capability while being completely flexible that avoids that lock and you’re able to customize and sort of expand upon it. It’s multi-vendor. Okay, so you’re able to see across the entire service chain which mitigates that isolation approach. Data gathering from all the devices really allows you to capture kind of that single source of truth of the network which is going to be super important to being able to draw analytics decisions.
So once you have the data the question becomes well what do you do with it? So the ATOM platform has automation is really the core component or the core capability for things like provisioning and sort of managing the network. But they also have the data there so it makes them a great place to be able to open up that data to analytics using what we talked about that modern sort of data science of machine learning. And what that allows you to do is really start to draw insights from the network based on the operating state. So a secondary approach is really around compliance and assurance and how you expect the Anuta network to be operating at that moment. So if you take a look at all of those areas of the network, that’s not a trivial set of functionality for a platform.
About the Massively Scalable Anuta ATOM platform:
[00::13:11] So one of the things that we talk with clients a lot about is just the platform itself. Because they expect if it’s going to be handling all of those things and it’s that volume of data and it’s that critical. Well, you better be rock solid underneath and if it better be easily integrated into other things that I have. So when you look at ATOM, it’s based on open standards. We talked about things like YANG for modeling. Popular programming languages that are dominating the industry like Python for extensibility. And so when you look at things like that, it starts to show the true power of the platform.
In addition to the idea of hey, we are talking about networks that have thousands of nodes potentially, have enormous amounts of data. How you are protecting a platform like this to be able to scale and sort of take advantage of things like. Well, the ATOM platform is built upon the same principles that modern applications are built upon. And they are using things like micro services, containers, message buses, the idea of stateless applications to really start to solve for this. So that really allows them to sort of under the hood to be able to expand out to meet a lot of those demands.
[00:14:19] Interviewer: That still issue is one of those things that just comes to mind as we talk about our network is getting larger, they are getting more complex even though everyone keeps telling us that they are getting simpler. We are seeing more and more data. And in that, you know we’ve already had issues in this scaling right when we talk about our monitoring platforms. I know I can picture them and I’m not going to name names of platforms that are just incredible difficult to get up to just a few hundred nodes. It’s really, really difficult. So I think this is one of those really strong things for Anuta specifically is the fact that they have built on this for microservices architecture. It will grow so long as you feed it and operate sources to whatever size it needs to go which is a really strong suit.
About The Future of Network Operations:
So we’ve spent a bit of time, now this is one of my favorite things, we’ve spent a lot of time talking about all these new things. And what everyone loves to do is to talk about all the things that are coming and all the fancy new technologies. We talked about quite a lot of applications and I like that. What does it mean for a network? So the next few years, like you talked about a lot of tools here that maybe networkers aren’t real familiar. And so in the next years, what do you think that role is going to look like and what should networkers be doing to get ready for what’s coming?
[00:15:24] Ryan: Wow, that’s a great question. Well, one thing I’ve learned certainly in the tech industry Jordan is that we are awful at predicting the future. I actually think I thought HD-DVD was going to beat blu-ray at some point. So I’m probably not the best person. So I’m a little skeptic how to predict anything too far in the future but I could probably hedge a few bets and mention things that are emerging now and likely to keep trending in the foreseeable future.
So let’s start with an easy one, right. Network bandwidth will continue to increase. I think that’s no surprise to anybody. I mean we are seeing creating option for a hundred gig. We are seeing 400 gigs next year. We are probably going to see 1.6 terabits very closely on the horizon. So to your point, network is certainly going to get bigger as far as bandwidth.
They are also going to spread out a lot more. So IOT and other edge computing are going to drag that network out to places we never dreamed it was going to be. Which puts a heavy emphasis on how you manage the network going forward. There is no mystery in the idea that we are seeing open devices. We are seeing increased automation. It’s nearly 2019 if you are not taking advantage of automation, you’re likely going to be left behind. So I think that’s a very real thing.
Another interesting area we’re seeing which I think is worth mentioning is just the contribution in terms of code and scripts to the open community through things like software repositories. Like we are seeing a huge amount of upswing in that. And I think that’s a great sign that we are seeing better collaboration in the industry. And people understanding that we’re all trying to solve the same problems. So like let’s stop isolating our solutions a little bit, start opening them up which I think is pretty much part and parcel for the open source community which I think is great. We have clients say, yeah automation is great but like the complexity of managing the network is there.
Jordan you made mention to it earlier like network isn’t getting any easier. So we are seeing things like intent-based programming or intent-based networking as like sort of the next shiny button that everybody is sort of looking at. And that’s really where we are going to be able to describe to an algorithm, the desired outcome or the desired end state of the network. And that algorithm is going to have a very keen idea of the running state. And it’s going to be able to deduce, here are the complex command sets that I need to substantiate in order to make the desired the running state. So I think we are going to see that capability working into several areas of networking.
[00:17:54] And I think the last thing really is like as the network matures we are going to see a tighter integration set of the network into that tech stack where networking analytics is going do in sight and we are going to see things like automation self-provisioning that’s really going to empower that programmatic approach. And really take us to that consumption model that we want which is hybrid cloud. So I think those are the main things and so like to your point, what can people do? I think people need to start connecting with each other and really start learning a lot more, opening the network up so that people are sort of able to use each other as examples for how we should be on ball in these things.
The Final Call:
[00:18:31] I love that answer. I don’t know if you know this or not, but that’s completely what network collective is based on. That whole idea of connecting network engineers, our motto is connecting people who connect the world. That’s really cool. So thanks Ryan for coming on especially to talk about your experience with Anuta. Thanks to all of you who spent the time listening to this short take. If you need to find out more about how Anuta can help you with network assurance and orchestration. You can head on over to anutanetworks.com/collective. There you will find lots of great information. Some white papers, case studies and you could even demo the platform for yourself. Thanks again to you all and we’ll see you next time.