Networking in the Multicloud Computing Era – Amir Khan, Alkira

This is Textron TV. Hey guys. Thanks for the throw.

We're here with Amir Khan is the CEO for alkira. We're talking about sd-wan sassy and what's going on with remote access and general Amir. Welcome the show.

Thank you so much Michael. Thanks for having me. I think when the pandemic initially hit we all thought that we'd be on some sort of sd-wan Slash sassy setup by now people would ditch their vpns and yet it's been you know, so progress it's not clear to me.

It's moving any faster than it would have its own normal Accord. But what's your sense of where are we in this migration? And maybe the journey is just longer than we all thought.

Yeah, so there are a couple of things that happened after pandemic people started working remotely from homes. So perhaps the branch transformation to SD ran slowed down a little bit. But you know at the same time the cloud adoption was increasing so people are trying to put more more into the cloud without shipping any hardware or any, you know devices which were being delayed by because of the manufacturing reasons.

So the cloud adoption started taking off and what we have noticed is that cloud infrastructure in our case, we build a common networking infrastructure across all clouds, and that that's taken off quite a bit. And what is the level of sophistication in networking right now and I'm asking the question because there are so many people working from home using a VPN to go to the branch office. And then from the branch office, they're shifting traffic back to some data center through some backhaul mechanism for the cloud access and the latency issues as you imagine probably aren't as fun as most users with like So, where are we in this adventure?

Will you how complicated is what we're calling Cloud networking these days and what are we doing wrong? Yeah, so first of all, you know users want to access where the applications are residing directly as quickly as possible, but the least amount of latency and the highest performance possible. The applications used to reside purely in the data centers and now depending on which customer you talk to.

It could be somewhere between 30 to 70 percent applications sitting in the cloud now so customers want to hop onto the cloud as quickly as possible. And what we allow them to do is by creating a common infrastructure across all hyperscalers. We allow them to come directly to the cloud as if they would have had access through the data centers into the cloud.

So basically we are shifting all the applications which you used to reside like firewalls like Cloud, you know ipam Services into the cloud so that the customers don't have to go to the data center and basically circle around and go back into the cloud. They can directly go into the cloud in a very secure manner. And if that's the case who's making that decision these days who seems to be at the Forefront of this transition because it seems like any kind of a dedicated networking team but not every company has those these days so what does it take to kind of get to this new era of networking and who leads the charge?

Yeah. So application development teams the cloud teams in in the customer base. They have the requirements they come up the requirements and then they work with the security and networking teams to come up with the solution and still primarily for the network piece Network team is responsible for the collaboration between teams is increasing as as the ship takes place to the cloud.

Do you think this new area of cloud native applications? Therefore is going to drive a lot of this transition because they are made up of these microservices that are very latency sensitive. Oh, well, what we have noticed is that microservices still reside in a single Cloud a majority of our customers are telling us the reason they adopting multiple clouds is because custard the the users are accessing either applications, you know different applications like directory services in one cloud and some servers in another Cloud Etc.

But it's not much of East-West traffic between clouds. So yes, the there is increase of the cloud utilization but we still don't see microservices spread across multiple clouds now within the clouds it's already optimized for microservices, right? So where you want to have the ability or the flexibility is in the hybrid environments the majority of the traffic from users weathering home or now with more more people going back to the offices.

They can directly access the cloud and majority of the traffic is north-south. There's very minimal traffic as we talk to the customers, which is going between clouds whether it's between Cloud applications or microservice Services being spread across the clouds. Each of the cloud service providers seem to have their own SD and network some type is why do I need something that sits above that?

Am I trying to centralize the management of networking across multiple clouds despite the lack of integration between the clouds is that the idea or is there something about the way those networks are constructed that they might lock me into a specific Cloud provide. That's a very interesting question. When when we embarked upon this journey in 2018 to build a common Cloud Network infrastructure.

We studied the clouds quite a bit. And what we would noticing is that Enterprise was Enterprises were struggling. Even adopting a single cloud and the slowest piece with a networking piece.

The applications were agile, you know, even the compute storage were quite agile as long as you were spinning them up in pieces in Virtual, you know private clouds Etc. But when you brought it together with the Enterprise environment networking environment, the segmentation was lacking the number of routes that you could carry across regions was lacking and to in policy was lacking end to en route control was lacking this so there were so many different details and then on top of that the cloud Network piece was stitched together manually and one of like our customers called it spaghetti mess because it was all brought up manually by stitching tunnels together between instances. So let's say you want to deploy a firewall you had to figure out how to do that.

If you wanted to distribute traffic across by our walls that make it last week. That was a problem. So what we notice was that there was no Enterprise class care class capability available inside the cloud networks for networking.

And their own infrastructure was very highly available. But what Enterprises need High availability for networking you had to do it on your own you have to deploy nodes and multiple zones. You had to figure it out as a customer yourself you had to figure out inter region redundancy disaster recovery.

So the poor customer was, you know left to learn one cloud, and then they were requirements coming in from teams to you know, adopt another Cloud because of optimization of certain applications Etc in or let's say you merge for the company as an example. Then your team had to learn that cloud every cloud is done different than another Cloud came into the picture. So on an average, they're more than three clouds that are customers are using and that's what we've noticed across the board, especially with large Enterprise's and that was becoming a nightmare for customers to handle and there was a dire need for a infrastructure like ours they were orchestration.

Space for us who were utilizing the underlying capabilities of the clouds to make it a little bit easier. So think of it as Network management capabilities of the past in the networking industry, but that was always lagging behind the cloud providers capabilities. So what we did was we built our Cloud inside the clouds.

We utilize the cloud capabilities and we built our infrastructure inside the clouds because we didn't want to have any dependency on Microsoft Azure puts out this feature in networking and AWS puts out this feature in networking. We wanted to provide a networking solution for the Enterprise which was Carrier class. And that's what we have built.

And once we did that now we had this common abstracted layer across all clouds, which seamlessly interconnects with the hybrid environments and then we integrated best of breed applications. Or services from Palo Alto Cisco checkpoint Fortinet whatnot, right all so that customers could standardize on whatever firewalls they were using. Whatever iPad Services they were using they could standardize that you know, I'll give you an example.

I was last week at the open network user group. One of our customers chart Industries presented there and they were struggling with iot traffic going all the way back to the data centers and tromboning into the cloud right and with our capabilities. Now you have shifted the you know, many data centers into regions.

And now all those applications can directly access the cloud in a very secure manner. So it's a very comprehensive solution which allows them to meet their compliance requirements the day to operations the troubleshooting which they were not able to do before that. You know, imagine if you have 10,000 tunnels, you know, interconnecting a thousand two thousand vpcs and the hybrid environment and the more users everybody coming in it becomes truly spaghetti mess.

And if you do any kind of day to operations and Upgrading all that software, right that becomes a nightmare. So what we have done is taken over that responsibility from the customers and providing a SAS type solution for networking for the first time in the history of computer networking. So that the customer doesn't have to worry about any of that.

I hope I made so myself clear. I'm very passionate about this stuff. It's it's really exciting.

No, I actually do you think at some point we'll be moving more to this hybrid cloud computing environment that people envisioned or will everything kind of still sit in this one, you know class of work with those will run in one Cloud but not in the other and these things will be remain some what semi isil. Well hybrid is the the thing, you know, I think it'll be a long time before applications move hundred percent into the cloud especially for large Enterprises. Now their companies like ours which are born in the cloud.

We you know have our own physical Plumbing done inside the branches and stuff like that. But beyond that everything resides in the cloud, we don't have data centers as startups, right but those are new age application companies which are coming up majority of the companies that have existed in the past for a long time. They they have to you know to to redo the applications in many ways to you know, the transform them into the cloud.

So they're doing a lot more now, we are starting to see Banks and you know hospitals and all these guys coming in now saying that we are starting to move more more applications into the club, but it takes time. So for the foreseeable future what I think is going to happen is that we will see hybrid environments for a long long time and that's why we Integrated our solution with the on-premise, you know capabilities that Cisco is in junipers and aristas of the world offer and then on the other side we connected to the cloud and we become that glue to allow these multiple environments to exist together, right and then as far as multiple clouds are concerned that is happening today. And as I alluded to it earlier customers are accessing either because of redundancy or merges of Acquisitions or certain applications running better and cheaper into a particular Cloud, but that whole thing needs to come together as a network segment for the users to consume them in a proper manner from the hybrid environments, right and that's what's we are noting across the note saying across the board, you mentioned developers and development teams driving this transition.

And of course devops is heavily used by those folks. Do you think network operations is gonna I'm an extension of devops as we go along here and are these things going to become more programmable? And this is how I'm going to approach that or will it be?

Still like graphical tool would dedicated network administrator or some combination of all those things. What's you know, what's your crystal ball telling you? Know devops is a way of doing things.

I mean there are tools available to do many many different things. Right and that networking happens to be one of those things and the way we look at it is that there will be more automation more apis available the more capabilities like terraform available which we provide. We are a terraform provider for our solution.

And you know the networking teams can program that now what happens is that you still need deep knowledge of networking to be able to control the network piece. Right? So, yes, the tools are coming together, but the teams have to learn individually all these capabilities so that they can create that devops type environment.

Hmm how automated can all of this get in the future. We hear everything about machine learning algorithms AI all kinds of magical things are coming down the pike. How smart can smart get when it comes to managing networking.

So that's always an interesting question because you know in networking as we started deploying these large networks even in the early 90s, right we used to deploy networks with maybe 50,000 nodes in many cases because they're used to be many more branches at that time, by the way, right and it was lower speeds and fees, but the scale of the environment was very very large and the way we created hierarchy and for being able to control that environment because network is a very Mission critical environment. Right, so you have to have certain level of predictability when your troubleshooting if something goes wrong. So now you can bring in all kinds of new ai-based tools and machine learning in certain aspects of those solutions to probably isolate the problems.

But when you're taking actions people still like to control them and have an understanding of what impact will it have on the overall environment? So as far as the ACT is concerned we are not there yet. But I think people are starting to experiment with the AI type constructs to see how we can intelligently start to understand things better and then probably propose something to the network administrators because it's a very complicated environment mind you, you know, if a large bank or some of these Financial institutes if they have a network go down now, you're losing hundreds of millions of dollars an hour.

Right in transactions and that is not a very good thing. So certain level of predictability and understanding of how things will progress right if we fix one thing. How is it going to affect the rest of the environment?

That's very this is inherently distributed environment. It's a very complicated environment since when you know since recently we have started to understand the distributed computing well enough networking from the beginning has been Loosely coupled distributed environment, and that's why it was so difficult. Right?

And on top of that is this Mission critical because all your applications run on on the network if network is down everything is down. So that's why we are very very careful. Yes.

We are starting to use some of the machine learning AI capabilities, but very carefully. It is often struck me that networking is incredibly complex and the more platforms you add the more complex it gets and yeah and yet we don't hear about you know, there are outages and there is down times but it's not what I would say catastrophic so networks are surprisingly resilient. So why is that and and or is this like a problem that's gonna explode or we getting better at the resiliency side of this equation and it's and but we still have issues.

Yeah, so we have learned over many many years how to make redundancy High availability how to build it into into the networking environments as we deploy them. Right and that was a fairly manual process. We had to go through literally design every aspect of it in a very manual fashion.

I want to deploy this many routers in the background. This is the type of national partial mesh. I need to create this is how all things need to be designed from the routing protocol perspective to create a level of hierarchy.

You know and what if my mpls environment goes down do I have a backup to internet or some other mechanisms like lease lines? Right? So so we built that high availability into the networking environments and that's why people did not hear that much about it.

But every once in a while you do here about it when networks are down, you know, it turns into and oh that application was down and sometimes in the background. It was a network which was down but over time and that's what we are trying to automate now. So every customer does not have to build redundancy to their own levels.

We need to have Solutions which are fully automated right and which have that level of intelligence built in so that the network doesn't go down. Is it multi AC is it multi region? Is it across providers?

Right if my back one on one provider is down which you know likelihood is very low but there are some customers who cannot tolerate more than two minutes. Right in a year of downtime. I mean five nines as I think around 17 minutes if I'm not mistaken, right but there are certain environments which cannot even tolerate that so you have to build it into into the design of whatever, you know your solution offers to the customer and that's why so Mission critical what we build.

I mean it's adoption cycle is a little long but once people customers adopted there's no way they're going to out of that environment because you know, it's very difficult to move into an environment if you think about Cisco why has Cisco been in the computer networking environments for the last 30 years and why are people not able to get it out? Because there's so many tools and so many, you know surrounding things built around that to support that infrastructure people will have to redo that if they adopt a new solution. The reason companies like ours come in to the picture startups is because we are solving a newer problems for the customers.

There was nothing that from the previous providers in the cloud networking environment within the cloud. And AWS and you know Azure and all these guys they've done a tremendous job as for his computer resources or stories or applications or other things are concerned but networking was not necessarily their expertise to that extent and that's where the expertise that we've had over the last 30 35 years marrying that with the cloud computing environment and understanding of that. They're very few people mind you who have transitioned from networking into the cloud right?

Majority of friends are retired, you know, so we continue to work on that and for that you need the knowledge of both environments to be able to offer a Enterprise class fully available solution to the customers. All right. Well, it's a brand new age in networking.

It's just a question of how long it's going to take us all to get there. Hey, I mean thanks for being on the show. Thank you so much for having me Michael.

It's a pleasure talking to you and please reach out anytime if you would like to discuss anything for. All right back to you guys in the studio.