Simplifying Kubernetes: From Edge to Cloud
Sign up for Daily CommTech Brief here.
Session Abstract:
What does it mean to be Application-aware with advanced data management capabilities to Kubernetes-native frameworks, for Cloud Native storage? This session will offer examples of extending agility, efficiency and portability of Kubernetes to applications like AI/ML and Custom Apps, on any Infrastructure, On-Premise, Hybrid Cloud or Multi-Cloud. On this session are, Robin.io’s Tushar Doshi, he is Senior Technical Director, we also have AT&T’s Rupesh Chokshi, VP of Product Strategy and Innovation and on the end is Intel’s Rajesh Gadiyar, Vice President and General Manager of the Network and Edge Group.
Executive Speakers:
Tushar Doshi - Senior Technical Director, Robin.io
Rupesh Chokshi - VP, Product Strategy and Innovation, AT&T
Rajesh Gadiyar - Vice President and General Manager, Network and Edge Group, Intel
Transcription
SIMPLIFYING KUBERNETES: FROM EDGE TO CLOUD
Abe Nejad: And what does it mean to be application aware with advanced data management capabilities that for Kubernetes native frameworks, for cloud native storage, this session will offer examples of extending agility, efficiency and portability of Kubernetes to applications like AI ML and custom apps on any infrastructure that's on-premise, hybrid, cloud and also multi-cloud on this session are Robin.io Tushar Doshi, he's senior technical director. We also have at AT&Ts Rupesh Chokshi, he's vice president of product strategy and innovation. And on the end, next is Intel’s Rajesh Gadiyar, he's vice president and general manager of the network and edge group and gentlemen.
Good to have you again, we've done this just before. Tushar this is new for us doing it for the first time. So it'll do a great job. And Rajesh we've done this several times before. I believe the last time was the last time we were all in Barcelona, which was 2019. So it's good to have you. So this is somewhat of a technical discussion as opposed to more of our high-level thought leadership discussions. So should be interesting. I want to start with you Rajesh, if you don't mind state full workloads, a little of a tongue twister, being deployed by containers or in containers by public infrastructure as a service providers. Why is this trend increasing?
Rajesh: Yeah, it's a great question. But before I answer that, first of all, it's like you said, great to be here in Barcelona at the mobile Congress, in person meeting with real people. And second, actually I want to start by congratulating Partha and Tushar and the Robin team on the recent acquisition. Congratulations. And so coming to your question, actually, so it's really interesting times. So there's been a really nice buzz about 5g and of course, 5g is a great technology. We've been talking about it for the last few years, but it's really what 5g affords. And if you look at the new-age services from autonomous driving to video services and immersive media, they all actually have a requirement for low latency and quality of experience. And as a result, it cannot be just transactional like it used to be, like a device talking to the cloud.
So many of these new age requirements really require the cloud to come closer to where the application is and where the data is sized. And so in a way you can think of it as basically a marriage of cloud and communications or compute and communications. And so when you sort of think about it in that partly is what becomes really important then is unlike the transactional nature of like what the device and cloud communication used to be, on the wireless side, communication side, there are actually requirements associated with mobility and roaming and as a result there is a need for some level of state to be maintained because when a user is moving, clearly there are actually transactions that are dependent on each other.
And so as a result, actually the data that's associated with the users applications, it requires a slightly different paradigm. And so many of these applications need to be stateful. And this is where Robin as really excelled in actually helping many of these applications that require that stateful, have the stateful nature. How do you actually manage the data efficiently? How do you actually have the metadata-aware application scheduling? So that's basically what's going on and why there is a need to support stateful and like managed data accordingly.
Tushar: So just to add to that, so in Kubernetes, it's mostly designed to run stateless applications to start with. So now people have realized the importance of Kubernetes for managing their applications, so that includes a healing or scaling or other workflows. So it essentially became that it's very easy to manage when you have a platform like Kubernetes supporting you. So now people have realized the importance of running stateless applications, and now they're thinking, oh, so I can deploy stateless applications there. Why not add all those benefits back to the stateful applications? So that's where I think people are looking to adopt more and more stateful applications in Kubernetes.
Rajesh: It's actually about, like Tushar said it's about ease of use. So you don't want actually to manage the stateless outside of Kubernetes. Kubernetes has become the tool of choice. So you want actually provide that ease of use by providing the ability to manage both stateless and stateful applications with the cloud-native approach using Kubernetes.
Abe Nejad: Again, for the audience out there that maybe not as astute on this subject, as you are, the correlation between being application aware and cloud-native storage, can you sort of tie those two together for us?
Tushar: Yeah. So before Kubernetes, people are used to managing applications in a particular way. So say my Sequel administrator or a Kafka administrator. So what I'm looking for is my application should be easy to deploy, but it's not just that the day zero I deploy and that's done. So what they're interested in is what happens in day 2, what happens if somebody makes a mistake and drops a table from database, how do I recover from that? So people are looking to have a support for say, snapshots, where you can go back in time using that, or people are looking for backup, which could be on premises or backing up into the cloud so that I can recover it whenever I want, or they want the cloud portability where I can move applications from one data center to the other and so on.
And of course, with the compliance and all coming into picture, the disaster recovery is also an important question where if I mean, we have seen examples that even public clouds are not resilient to site failures. So you have to have, your business should be running no matter if the site goes down or not. So now if you consider all these use cases, people want these use cases, and this should not be just, this is not just the storage problem or network problem or the compute problem, it's application problem. So what we have realized early in our design and development is if we have to provide all these lifecycle capabilities, it should be application centric and not just the storage compute network centric.
So that's where the application awareness comes in. So now when you take a snapshot, it should be the whole application, which should be snapshoted or backed up and so on and not just the part of it. So that's where the application awareness comes in. And in the cloud-native storage, I mean, when you are talking about, since we talked about that stateful applications are now coming to Kubernetes, and they're important for 5g in terms of subscriber management or the PCF, the policy core functions. So it becomes very important to have the similar capabilities at the application level rather than the individual piece level. So that's why it's very important to be application awareness when people are talking about onboarding them for those use cases.
Abe Nejad: Rupesh from an operator's perspective of the correlation between application awareness and maybe just cloud native.
Rupesh: Yeah. Yep. Absolutely. And I think we, early on got on the journey of software defined network and application awareness was the key. You make the network aware of what the application wants to do with the investments that Rajesh was mentioning in terms of 5g it's all here. So when you have a high speed, intelligent app aware software defined network capability, the modernization of the application, the ability for the use cases that you were describing Tushar is very important. So I feel that the capabilities are all there and there is an opportunity to make it even more better.
You mentioned about recovery, et cetera. And I would say in the world of cybersecurity, just the ability to rebuild, get more operational again, having distribution at so many different places requires a lot of technical juice. And kind of think about it very differently.
Rajesh: Just for illustration. So I think like you know, application aware, if you sort think about, there are applications that require maybe strong encryption, others might actually require maybe faster data transfers and not such strong encryption. Some application may require data compression. There's probably a need to sort of coordinate data as well. So this is where actually the nature of some of these applications really require application aware, metadata aware kind of approaches. We just want to actually give you a couple of examples of where this thing really comes to the floor.
Ade Nejad: Absolutely. I was just going to say Rupesh if you could offer maybe an example of extending agility, efficiency, or portability of Kubernetes to these applications like AI ML and custom apps as well. Is there an example?
Rupesh: Again, we are here at Mobile World Congress, here in Barcelona, you walk around, you will see tremendous amount of computer vision. You see cameras, videos, lot of data being captured and being analyzed at that sort of localized edge and applying the AI models to figure out what do I do with it? Like a lot of use cases that I saw was that, Hey, can we do fault prevention? Is the robot working at the speed that it is supposed to work? What is the efficiency that you're going to gain in the hospital when you do remote patient monitoring or trying to recognize how the patient is reacting? So the amount of technical capabilities that you need to do all of these things in real time, whether it is the AI models or the machine learnings or the localized processing, I think is going to be equally important to deliver these use cases at scale
Abe Nejad: Tushar an example.
Tushar: Yeah. So I think I just wanted to go back to the point of cyber security. That was a major point. I mean the instances that we have seen is the ransomware cases where, I mean, the data is encrypted. Now you have no way to go. And I think one of the example, I don't remember the exact, but there was a hospital data, which is very critical. Now, if you don't have a backup then it's very hard that you could survive your business. So that's very important. Coming back to the question of agility, efficiency and portability. So now as we have seen, so what people like is as much as vendor agnostic as possible, whether it's cloud environment on premises or so on.
So we have seen examples where even 5g it's not just the on premises, but even the clouds are offering are trying to come into the same space. So what people want is the portability of their application, whether it is cloud, on premises or any other any other vendor. So that becomes very important that you provide portability for these applications independent of what underlying platform is. In terms of efficiency, I mean, 5g, as you know, it's very latency sensitive. All the applications are very latency sensitive. That's where I think AT&T and Intel has a very important part to play, to come up with the infrastructure and the ecosystem for achieving this higher agility and efficiency.
And same thing goes for the storage where you have the application running and you have the subscriber data, and you don't want any latency just because your storage is slow. And that's where what we thought is, okay, we have to have a focus for running this latency sensitive applications. And that's where the importance of all this feature comes in the applications that are really latency.
Rupesh: One point I want to add, from a kind of a simple layman's perspective. The way I view it is that Kubernetes is providing a platform that you can easily port. So you're not trying to create an application fit for use that says, I'm going to run it over here. So I got to create it this way. We know that customers and developers are going to be in multi-cloud environments. So their ability to port those applications, utilizing the Kubernetes platform, gives them tremendous advantage. Otherwise, you are stuck where you are, and then if you need to make significant upgrades or deal with security issues, et cetera, the amount of time it's going to take you as a developer and the owner of the application is significant.
Rajesh: Yeah. And maybe the other thing to think about also is automation at scale. I mean, we are dealing with deployments that run into maybe thousands of cameras. And thousands of edge locations. And all of these actually you can no longer roll a truck and the ability to sort of like automated scale, bring up these platforms. I think Rupesh mentioned analytics, the ability to sort of like, take these thousand streams and look for certain kinds of patterns in an automated fashion, it's becoming increasingly so critical. All of this actually would require low latency processing. And all of this would need to be application aware.
Abe Nejad: I wanted to ask you about, and we can go down the line here about hybrid cloud capabilities and why they're so important for let's say applications like cloning an application. Can you talk about more about that?
Tushar: Yeah, so when we were in journey, when we are trying to get this application at the production level on Kubernetes, what we saw was, there are typical challenges that even non Kubernetes applications also see. So one of the challenges I'll give you an example of, say, I want to upgrade my application. Now upgrade is a cumbersome process, no matter how you have your application design. So we said that, okay, you have this problem. How do we solve it, since Kubernetes is very flexible and agile environment? So the thing that we came up with, okay, what if I can test my upgrade before actually performing an upgrade on your production application? So the way we solve the problem is, okay, your application is running you take a snapshot. You create a clone of that application. So essentially you have another copy of the application running, which has exactly the same state as your production application and run upgrade on your clone.
So now you are not impacting your original application, but still you have tested your upgrade good enough on your clone. If that fails, you fix it, try it again and you repeat the process. When you are satisfied, you go back and upgrade your application. And same thing. I think we talked about cloud portability. That's one of the powerful cloud capabilities that we need in the environment where things can move around, whether it is part scaling, like say you have a holiday period and say, you need more resources, which your application typically runs on, on premises, but for holiday period, you need that burst.
So now you say, okay, I want to put my application during, Thanksgiving or Christmas period when people are actually doing a lot of shopping and deploy it in the cloud for that time. You scale as much as you want, when you are done you go back to your on premises. So that's one of the use cases that becomes important and same thing, is what we have seen is the VR is very important for these application nowadays, because one is you don't want your business to be down no matter what, plus another is from compliance perspective, you might have to do it as well.
So now when people are talking about VR they're saying, okay, what if I have one site which is on premises, I don't want to invest in multiple on-premise sites, so I can have a cloud as one of my target location. And I can have that functionality where I have a cluster where the application is running on, on premises, and I do VR on the cloud environment. And that becomes one of the powerful capabilities that you might need in your applications and your platform.
Abe Nejad: Rajesh I want to talk about partnerships and collaborations. Let's talk about specifically bringing advanced data management really to power stateful applications. Can you talk more about that?
Rajesh: Yeah. I think what better example than actually the two gentlemen that are sitting here. So we've been actually collaborating with Robin for a long time in this area. And I think the level of automation that actually they have delivered across hybrid cloud environments and at scale for 5g in the radio access network. And these are real deployments that scale thousands of nodes in Rakuten for example. AT&T has actually done similarly, really massive scale automation. So I think from a partnership perspective what Intel does really well is how do we sort of like lift all boards and really sort of like grow the ecosystem. So that's been a lot of our focus.
The other partnership, which probably gets less attention, is like really what happens behind the scenes actually, in terms of how we collaborate in open source communities. I think a lot of like what we're talking about here in terms of application aware, data management and especially in the context of cloud native and Kubernetes. Kubernetes actually affords us, what's called a container storage interface CSI. And a lot of like the nature of this data that we are handling, it's actually persistent data. So what Intel has done is a lot of focus on, of course performance. And how do we actually improve the compute performance for processing databases. We also brought in a technology for process memory, and then what you've done is actually developed a container storage interface for what you call a PMM, A CSI that actually makes the process memory look like basically a file system.
We've been actually working with Robin on this one as well. So these are the kind of things that we've been actually looking to do both in terms of our partnership and communities and open source communities as an example, and also in the ecosystem really driving what is needed to really deliver the word use of distributor processing 5g and edge, and then partnering with companies like Robin and AT&T to really sort of like take it to deployment.
Abe Nejad: Before we get to Tushar, is there anything you want to add to that?
Rupesh: Just to build upon what the gentlemen said. So one is, I like what you said, where the innovation is happening up and down the stack. So whether you are a chip manufacturer, whether you are a service provider, whether you are a software technology provider and we're sort of building upon each other. So you put this interesting demand. So the service provider says, Hey, I need to do this thing at scale. So I need more kind of processing power and I need the software to be faster. And then vice versa it goes up up and says, Hey, if we can put so much power into a device or a handset, what more can we do, et cetera?
So that is one point, which I think is excellent on the innovation side. The second is I think we live in a world hybrid. I think just like in real estate, people talk about location, location, location, I think today, and for the next, so many years, it's going to be hybrid cloud, hybrid networks, hybrid devices, you name it. It's going to be hybrid hybrid hybrid and giving the ability to kind of go back and forth, which is what some of the software platforms that we're talking about provide is very important, very important.
Abe Nejad: Tushar advanced data management, to power state full applications. What's your perspective on that? As far as collaboration and partnerships.
Tushar: So as I think Rajesh was saying, Intel has been a phenomenal partner for us, and I think the open-source contribution that you guys have in terms of multiple plugins and various CNI has been, I think the most important piece in the puzzle that was needed for bringing this 5g applications onto Kubernetes. So thanks for that. So along with other partners, what we are working is we are also trying to see if our storage stack, so even independent of what Kubernetes platform is, whether we can say OEM for particular appliances or their infrastructure. That's one. The other one is we are also collaborating with partners like Altiostar, and [20:50 inaudible] and other guys, to see if we can qualify their applications on our platform and have whatever the requirements that we have that they have, we will solve those so that they are essentially certified to run on our platform.
And that's where I think the partners who are the application, whether it's a ran, core or edge application, we work with them and make sure that they have what they need to run our platform. So that's what the level of partnership that we are bringing into the community.
Abe Nejad: So the title of this session is simplifying Kubernetes data management from edge to cloud and Rupesh, you brought something up interesting that wasn't on this session, it was on a previous session. You said the word simplify is such an important word, and you really have to qualify that word so people understand why it's important. Can you give us some insight into that and how it applies to this system?
Rupesh: I think it applies over here too, because we're talking about some very sort of complex technical things and capabilities that have to come together. And I feel that, the good news is that we are back again, more in person. So the learning opportunities are better, but in order for a lot of what we are talking to scale, the adoption to significantly accelerate simplification. Rajesh talked about some things, snap-in, they fit together better. The ability to do the portability, have a very efficient methodology. All of these things drive towards the simplicity that we need. Otherwise it's going to take us a long time. And will we keep talking about the same topic? We want to move past the talk and walk the walk.
Rajesh: Yeah. I really like that. I think simplify asyou say is the key tenant of what is required in the industry today. And the way to do that in my mind is like three things, composable, make everything composable. Make everything automated at scale. And then I think a lot of it is about heterogeneous computing as well, different kinds of resources. And how do you actually do the processing at the right locations? So you can actually meet the quality and deliver the quality that's required for the applications that we are trying to actually enable in the industry.
Abe Nejad: Interesting Tushar anything?
Tushar: Yeah. So for simplification, as I think both of you have rightly mentioned. So very important point and so just deployment or management of application is not enough, but what you need is a higher level, single pane of glass, which is managing these thousands of cluster applications. And you need to have right observability and monitoring platform and insights in terms of alerts events and all that. That is very important for this piece. And with the partnership and with the qualification that we can do, we can make it very simple for the end user, which is, I think the telco provider, which essentially is our goal to make it very simple, to deploy, manage, and actually observe.
Rupesh: Yeah. I think the world is autonomous, everything.
Abe Nejad: That's right. Tushar it was again, like I said, at the top of the program it's good to have you, you're a wealth of knowledge by the way. So it's always good to have people like that on these discussions, especially when it gets somewhat technical and you're clearly a technical person. So good to have you and I'm sure we'll have you again sometimes soon. Rupesh you've given very generous with your time while we've been here in Barcelona as always. So we appreciate that. And again, great perspective from the operators point of view.
Rupesh: Thank you for having me.
Abe Nejad: And Rajesh people are thanking, you left and right for your contributions on many fronts. Didn't surprise me. I've known you for years. And I know that you're in a lot of ways, the bedrock of what happens at shows like this. Maybe behind the scenes a little bit, but I know you, so I know that you're there.
Rajesh: Yeah, it's a real pleasure to be working with all our partners. And more importantly, it's actually a real pleasure to be here in person meeting with real people. I've been missing that so much for the last couple of years. And then thank you for the opportunity to be on this panel. It's been great.
Absolutely.
For any inquiries, please email anejad@thenetworkmediagroup.com