Earlier this year, Mirantis VP of Marketing John Allwright had the privilege of sitting down with Mayank Kapoor, VP of Engineering at Reliance Jio and Jay Lyman, Principal Analyst at 451 Research, for a lively discussion about public, private and hybrid/multi-clouds, and the role of each for companies in different industries and stages of maturity.
Watch our webinar recording and follow along with our transcript to hear how Reliance Jio rapidly grew from a tech startup to the largest telecom in India, serving 400 million paying subscribers with a combination of OpenStack provisioned on 4000 bare metal nodes and Kubernetes serving as a unifying layer for hybrid cloud deployments.
Jay Lyman 0:08
Hello, and welcome to our webcast: Why the World Still Needs Private Clouds, the why and how of going cloud native with Kubernetes and
OpenStack on premises. I’m Jay Lyman, and I’m pleased to be joined by some experts in the field today. We’ll be going through some market data to set the scene, we’ll get some perspective, some end user perspective, and then we’ll have a hopefully lively discussion about private cloud and why it matters.
Jay Lyman 1:10
Just a brief introduction of our speakers. I’m Jay Lyman, Principal Analyst with 451 Research, which is a part of S&P Global Market Intelligence. I’ll give John and Mayank a brief opportunity to introduce themselves now as well.
John Allwright 1:30
Thanks, Jay. My name is John Allwright. I’m the VP of Marketing at Mirantis. If you don’t know us, we’ve been around a while, took a very large leadership role in the OpenStack community, and are doing a lot with Kubernetes after acquiring the Docker Enterprise business. I’m happy to be here and looking forward to the conversation.
Mayank Kapoor 1:54
Hi, Jay. Hi, John. This is Mayank. I’m a VP of Engineering at Reliance Jio, and for people who don’t know, Reliance Jio is a very young startup in India. We are only six years old now. We launched four years ago. But we are now the largest telecom company in India. So we have grown really fast.
Jay Lyman 2:18
Great, gentlemen. We’re glad to have you. Before we get into our panel, I’ll go over the agenda, I’m going to set the table with some market data from some of our survey work. And then we’ll get some perspective on things from Mayank. And then we’ll have our roundtable discussion.
Primary Environments for DevOps Implementations
Jay Lyman 2:50
So just talking about some market trends and perspectives that kind of highlight what’s going on with a private cloud, in the context of public cloud, hybrid cloud and hybrid IT, and also multi-cloud.
When we asked organizations about their primary environment for DevOps, as you can see, on premises private cloud is the leading environment and we’ll also see hosted private cloud, fairly heavily used, and also on premises non-cloud environments. Despite the fact that public cloud gets most of the attention, and it is true that organizations are increasingly relying on public cloud and SaaS services, on premises and private clouds remain relevant for applications and data that companies want kept behind the firewall. In many cases, this may be around sensitive data, or compliance and regulatory concerns, but the point is, that on premise and private cloud aren’t going away. Instead, we see the rise of hybrid cloud deployments.
Another thing that’s giving these environments staying power is the fact that a lot of CI/CD and DevOps release processes have been created in-house, ad hoc, and in a custom manner over the years and over the acquisitions. So, they’re somewhat beholden to on premise and private cloud environments. That is changing, as we see more reliance on public cloud and SaaS services, but again, that’s why hybrid cloud is so significant, and we think, here to stay for the foreseeable future.
Cloud/Hosted Services in Use and in Plan
Considering the cloud and hosted services in use or planned really illustrates the mix of tools, infrastructures, and environments in-play with today’s enterprise and organizations. No one seems to be all in on any one strategy, they are using public cloud services, they are using SaaS, but they are using on premises and private cloud tools that they like and know. They’re using open source software. So, there’s this real range of hybrid IT as well as hybrid cloud, meaning the infrastructure that they are running on. And we see use of private clouds, particularly hosted private clouds, which rivals public clouds in terms of planned use.
We also see digital transformation leaders, manufacturing, finance, all focused on IaaS, as in public cloud, but those organizations are still also using on premise private clouds and hosted private clouds. The focus is often extended beyond cloud migration and modernization, treating on prem and private cloud infrastructure and applications like they are in the cloud. We increasingly see, organizations striving to have a consistent experience, across on prem, in private cloud and public cloud. And so anything they can do, to make the look and feel for developers, for users, or even customers, in some cases — the more continuity that they get, it helps them deal with some of the complexities, especially when you get to scale. We also see giving the cloud native treatment to on premise and private cloud applications. So, the bottom line is that cloud native is not necessarily limited to the cloud, we do see those technologies, whether it’s containers and Kubernetes, and methodologies like microservices being applied to some of that on premises and private cloud application portfolio.
Primary Environments for Kubernetes
And here, we see with Kubernetes highlights, again, private cloud is a primary venue to run Kubernetes. But we also see, and this has been consistent over the last few years, use also on prem and in public cloud. Of course, we do expect that to grow, but I think Kubernetes in addition to being a container management and orchestration software, is a distributed application platform, and that’s a match for hybrid cloud. The same way that Docker containers matched enterprise need to package and deploy applications in the cloud, Kubernetes is fitting the bill for deployed applications across hybrid and multi cloud.
And we see the start of multi-cloud. We expect that to grow as well, and it used to be something customers wanted to have and now we see them actually leveraging multiple public clouds. This isn’t necessarily for one application, but particularly across their portfolio, all their applications and teams.
Factors Influencing Workload Venue Selection
There’s different reasons, and we’ll get into that next, the factors that influence this. Again, this is across teams and application portfolios, and it may be about different cloud capabilities with the different public clouds that are available. It may be on factors like security, cost, reliability, performance, data sovereignty, and regulation, and compliance concerns are always big. There may be organizational drivers. Another one that isn’t even on here is geographic location. But this is true for hybrid and true for multi cloud deployments, and it’s basically what we call a “best execution venue” approach. People aren’t using all these different infrastructures to make IT operators’ lives hell, they’re doing it because there’s reasons that some applications, some workloads, are better off being run on particular clouds.
Current State of Hybrid Implementation
This is the current state of hybrid IT implementation, and toward the top are areas where hybrid is mostly meeting expectations: faster application scaling, application lifecycle management, centralized data repositories, cost optimization, tying back to some of those factors that we mentioned. We see a little more work required. Things like portability, and bringing disparate IT environments together for more consistent and effective management. And this is where there’s, still some challenges in creating that consistency, across all these different environments.
We’ve had tool sprawl, we’ve had config sprawl, and now I think sometimes we can get into environments sprawl, but this is where it’s key for organizations to have observability and be able to see what’s going on and know how they can optimize and how to execute a “best execution venue” approach, and tracking these things. So, organizations have learned, and they’re getting better at it, but it’s certainly not mastered yet. We do see the industry on this road of hybrid and multi-cloud for quite some time.
Jay Lyman 10:43
So with that, I’m going to turn it over to Mayank to transition into our panel discussion by talking a little bit about the Jio experience and perspective on private clouds, hybrid and more.
Reliance Jio: A New Age Telecom Company
Mayank Kapoor 11:03
Thanks, Jay. Let me give our viewers a quick overview of Reliance Jio, especially for people who don’t know what Reliance Jio is.
Reliance Jio is a new age telecom company. We don’t consider ourselves a telecom, we say we are a tech company, and we are one of the fastest growing tech companies in the world. Since we launched, four years ago, until now, we have grown to 400 million paying subscribers, which is literally the fastest growth rate that anyone has ever seen. We did multiple smart things, multiple innovations along the way.
When we were launching, we were a default 4G LTE network, we had no 2G or 3G telecom bands to fall back on. There were no phones in India at that time that supported VoLTE. So we launched our own brand of smartphones, Life smartphones, very good value smartphones, that made other OEMs, other manufacturers, also launch VoLTE smartphones in India.
We made multiple firsts in the industry. We made voice calls totally free, anywhere in India, roaming charges to zero. We offer the lowest data rates in the world. Later on, we launched our own smart feature phone, because in India, there are still 500 million feature phones being used, and we made it free for our customers to use. That has helped us grow really fast, and we are now 400 million strong.
So, as I was saying, there were a lot of industry firsts that Jio did based on the technologies we chose, our cloud footprint. We enabled activation of a Jio SIM within 15 minutes compared to the previous state where it took 48 hours to activate a Jio SIM. So now, it’s possible for people to walk into a Jio store, give their Aadhar number, which is a phenomenal change, a systemic change, in India, a digital identity in India, and verify their fingerprint through e-KYC. Within 15 minutes, a Jio SIM is activated, and all the data that they had on their old phone is transferred to their new VoLTE smartphones.
We also made all voice free. That was the first for any telecom company in India. In fact, telecom companies in India were launched largely voice-revenue driven, before Jio. We essentially came in and said, unfortunately, that will not continue, we are making voice free, let’s compete on data, and then we made data the lowest price in the world. We still have the lowest data rates in the world in India.
We, again innovated, and we offer a full suite of cloud services, for free. You can watch TV on your Jio SIM for free, on your Jio phone for free, you can listen to music on the Jio phone and your Jio SIM for free. All the services we bundled, and we offer for free, and most of these services are running on our own private cloud that we will talk more about.
And last, I think it’s clear, in India at least the impact we’ve had which I’m very happy about. So before Jio, India was 150th in the world in terms of broadband internet data consumption. Now, we’re the first country in the world. Our users routinely consume 15 gigabytes of data per month on average, and it is forecasted to grow to 25 gigabytes of data per month. So it has been a phenomenal story and I’ve been lucky to be part of the Jio journey until now. So that’s a quick overview on Jio, and the impact Jio has had in India.
Running Apps on Both Public and Private Clouds
John Allwright 15:11
So I’m going to jump in there and take you back on your slides, because I think it’s very appropriate to the discussion we’re having, and I love that you think of yourselves as a tech company, rather than a telco because, what is it like building and growing the infrastructure that’s needed to support this kind of curve? How do you approach that?
Mayank Kapoor 15:36
Yeah, it’s been crazy, but it’s been full of learning. We are really heavy on private clouds. Even though we have applications running on public cloud, we have applications running on both private and public cloud together, the hybrid cloud, but we lean heavily on the private cloud, mainly because none of the public cloud providers in India had the scale to support these types of numbers. When we were launching, public clouds in India had maybe one region in India, where we now have about 30 regions inside India. All our own data centers. So, we were forced into building all this data center footprint ourselves and knowing everything, from security, compliance certifications of our data centers, to even certifications for our cloud. We went the full journey. It’s been crazy, but it’s phenomenal to see that happen, and run production workloads for 400 million consumers on the private cloud.
John Allwright 16:53
So you have the public cloud regions now. How do you think about cloud now that you have all the options available to you, which to use when? I mean, that’s the $64,000 question in this webinar — what do you put where, and how do you decide?
Mayank Kapoor 17:13
We don’t, in fact, we are very competitive. So I run the internal cloud platforms team, the engineering part of it, and we actually cannot force the internal teams to go into the private cloud. It’s completely driven by the teams. So if the team says no, I want to use monitoring-as-a-service or the RDS database, they have the option to go on AWS or Azure, because we have a significant partnership with Microsoft Azure now. So they’re never forced to go into the private cloud.
The challenge for us at Jio, is obviously there are economies of scale and cost advantages on the private cloud. So how do you give our teams the same experience, at least a similar experience as they have on the public cloud? We are relying heavily on Kubernetes for that, and we are building the additional path layer of providing managed services, database-as-a-service, monitoring service on top of the private cloud.
So the short answer to your question is we actually don’t decide for our teams which cloud to use, the teams are free to decide, but now the private cloud platform is competitive enough, where teams are deciding, hey, I think private cloud makes more sense for me, especially beyond a certain scale, this is the right choice. So, it’s very team driven, bottoms-up driven, on where to host their workloads.
Impact on Development Teams
Jay Lyman 18:48
Well, fittingly, that is completely consistent with our research, and you’re saying the same thing that our survey respondents do. But there’s a couple of interesting things there: one, the competition, it’s a choice of the developers and I think that’s testament to how much influence developers now have, rather than having the framework with the platform and the environment imposed on them, it’s up to them. If they choose something, it’s not necessarily shadow IT that the organization isn’t aware of, but it is up to them. That means that they’re going to be using the tools that they want, that they like, that help them get their job done quicker, and that’s going to drive efficiency and developer happiness and a focus on innovation, rather than fighting the organization with a platform that you don’t like. So my next challenge is to compete with the public cloud services.
But he also indicated there that that is what’s forcing the developer experience to be something like the public cloud, right? You don’t want to go into something and it’s like, gosh, they use clunky old stuff, you know, cobwebs up in the server room and, you know, what do they use around here? If you’re using Kubernetes, that’s the latest IT operations paradigm.
So the other interesting thing is that you’re a newer company, and I think there’s a lot of assumption that a newer company is going to be all public cloud, but you highlight that there’s still private cloud advantages for you. Even though you’re starting new, you don’t have this legacy, traditional IT baggage. Private Cloud is a forward looking move for you, and I think that’s interesting.
John Allwright 20:48
Yeah, it’s interesting, I think private cloud has been unfairly associated with on prem, the two things do go together, but then when you think on prem, people think more about historical developer experience, about raising a ticket for it to spin up a VM, it comes back in two or three weeks. And hey, isn’t public cloud cool, because I can use my credit card and they get one instantly. Actually, the two experiences, on a par, it sounds like if you get the experience right on prem, developers will like that just as much. And, we’ll come to it in a bit, but there’s probably some cost advantages as well over the public cloud.
From All-You-Can-Eat to Lift-and-Shift
Absolutely. In fact, it’s that all-you-can-eat buffet model, right? So initial teams start out having small, cloud builds, but soon they see oh, shit, my bandwidth costs are increasing, oh, I have to pay for load balancers right? So, then they say, but in the private cloud, I don’t have to pay for bandwidth, I don’t have to pay for load balancers, the IPs are also free right? So they soon realize that hey, okay, I’m getting almost all the functionality of a public cloud with an all-you-can-eat model, I don’t have to start worrying about these additional costs that are there. So yeah, for cost reasons, it definitely makes more sense beyond a certain scale.
John Allwright 22:09
Interesting. So do you reflect the cost back, do you kind of cross-charge back to the units that are using the infrastructure?
Mayank Kapoor 22:18
Yes, we do. So we benchmark our prices with the public clouds. It’s very challenging to get the data center expenses to the exact team, but we managed to get there. Still, bandwidth we don’t charge our teams for, there’s no bandwidth charges associated to their expenses on the private cloud. And then the load balances are a shared CapEx expense. So, it’s a CapEx model. So even though we try and reflect it back, it still turns out much cheaper for the teams.
John Allwright 22:59
Is that something you see, Jay? I mean, everyone assumes that CapEx/OpEx is going to be cheaper to move to a public cloud, but I don’t know if that is necessarily what people are experiencing now that they’re making the move?
Jay Lyman 23:14
Yeah, I think there has been that, “Oh, you know what?!” moment for a lot of organizations and the shock of the all-you-can-eat when you’re not eating that much or you left something on over the weekend. I think the public cloud fever has broken, right, there’s not this drive that digital transformation means throw everything into the cloud.
We’re seeing the idea that you can “lift-and-shift”, and get away with doing that without too much trouble, is giving way to the idea that maybe we should just treat our on prem and private cloud deployments like they’re clouds, right, cloud-native. We can include cloud containers and Kubernetes in our on prem and private cloud strategy. So, like I showed, private clouds are the leading deployment venue for Kubernetes. So, we see that while putting everything in the public cloud, there is now much more pragmatism going into how and where you put your workloads and a realization that, yeah, we want to keep our on prem and private cloud, and by the way, this might mean that we keep a lot of our tech talent, right and that’s hard to find. So we’re probably better off bringing along our teams with some of that infrastructure that’s in the private cloud, and still aligning it with that public cloud experience, but not necessarily moving everything to the public cloud because that can get expensive, and these other venues have remained competitive.
John Allwright 25:05
Something we’ve seen is this kind of repatriation. Some of our customers find that once they put a workload out in the public cloud, they get a sense for the capacity planning, the parameters of that, and once you get to that point, you know how you need to size that app, and the expansibility of the public cloud loses its value because your workload is there, you kind of know the parameters, and they actually bring it back on prem. So they’ve used the public cloud as a sizing exercise and then, they can really nail down the cost when they bring it back on prem.
Mayank Kapoor 25:43
Absolutely, that’s exactly what I see. So when the teams design their sizing, they actually have to think of peak capacity, which happens every day. So most teams don’t really worry about auto-scaling that much, even though they should and it’s the most efficient design and architecture. Most teams would size for the 95th capacity requirement and would keep it that way. It makes it worse. If you’re already sizing in the 95th capacity, then why not just do it on the private cloud? Absolutely. That’s exactly what I’m seeing.
Multi-Cloud v Hybrid Cloud
John Allwright 26:23
That’s great to know, it’s not just me. So let’s shift gears slightly. So we’ve talked about multi-cloud, hybrid cloud, and we’ve started off kind of talking about it like it’s an “all”. You know, you either put it on prem, or you put it in the cloud. But I’m interested in your journey, Mayank, and your definition of multi-cloud versus hybrid cloud and what’s important for each of those. Who wants to start?
Jay Lyman 26:54
Well, I’m the analyst, so I’ll give a definition, and it’s pretty simple. Hybrid cloud basically refers to use of on prem infrastructure with public cloud. So this is that best execution venue, you have your on prem, and your public cloud, and then you’ve got your different factors for going forward.
We haven’t talked, really, about a whole lot about the different public cloud capabilities, depending on how deep you want to go on and artificial intelligence, machine learning, or maybe it’s data compliance capabilities, these kinds of things, or security certifications that some cloud vendors might have. There are these different capabilities.
When we say multi-cloud, we’re usually talking about multiple public clouds, so this is where I think there’s a little more skepticism, right? People can understand that you’re using on prem and private cloud and public cloud infrastructure alongside each other, but why would anyone want to use more than one public cloud? Our research shows that they do have a primary public cloud provider. And as I said, multi-cloud used to be a “want to have”, but don’t necessarily use.
Now I think as organizations are learning, as they’re optimizing, how they go about placing their workloads and their execution venue selection, they’re finding that there are these different capabilities with different public clouds that make sense. And it might still come back to data sovereignty, or geographic location, or compliance and regulatory issues. Multi-cloud is basically more than one public cloud, hybrid cloud is different types of infrastructure altogether.
Reliance Jio: Navigating Hybrid and Multi-Cloud
John Allwright 28:55
So Mayank, who has been on the journey, how was your experience navigating through this?
Mayank Kapoor 29:02
In fact, let me give you some lower level implementation detail, which might crystallize hybrid cloud and multi-cloud for our viewers. So hybrid cloud is multiple private cloud and public cloud together, running the same workload. What we’ve seen is public cloud providers are able to give us public cloud resources and subnets and networks that are BGP peered to our own networks and subnets. So essentially, even though some virtual machines are running on the public cloud and some virtual machines are running on our own private cloud, they don’t feel the difference because the IPs talk to each other. Even though those IPs are run, one of them is on the public cloud and one of them is on the private cloud. So for us, hybrid cloud is a reality. All of our public cloud providers have peered the networks with our own internal networks, and our workloads and virtual machines can talk to each other without any latencies. That is a significant deal for us. So hybrid cloud, it’s a no brainer.
Most of the public cloud providers will do that for you. They will connect their networks to your networks. So, it’s easy to implement with the public libraries and that’s what I recommend as well. So it becomes essentially an extension of your private cloud.
Multi-cloud is the emerging trend, so, here’s how I define multi-cloud, very similar to what Jay said, but I say, a single team, running some workloads on private cloud, maybe a couple of workloads on AWS, and maybe a specific workload on Google Cloud. The same team using three different clouds. That, for me, is multi-cloud.
For example, we have one of our teams, who runs a specific image recognition workload on Google Cloud. But just that. They get the responses for that workload, and then they store it in their own databases running on the private cloud. So they’re essentially using multiple clouds for the same app. What is more common is that a team decides that three of my apps are on private cloud, but a few of these apps, because they need managed services and the team has thought that the app needs a specific service of that public cloud, I will run that app on the public cloud. That is also fairly common. And for us, Kubernetes has become that unifying layer, that standard cloud API on all these cloud providers and the private cloud. So it’s very easy for us to say “Okay, this next step, I will run on Azure, because they also offer Kubernetes.” It’s very easy for us to do now.
Reliance Jio: Cloud Choice
Jay Lyman 32:02
Would you consider one of your public cloud providers the primary?
Mayank Kapoor 32:12
So, specifically in Jio’s case: yes. We have a partnership with one of the public cloud providers, a significant partnership. If we had to choose between public cloud providers, and if your workload is fairly generic, nothing special that you need on a specific public cloud, yes, we do consider one public cloud provider as our primary, and that’s what teams are directed to use.
Right, but not limited to it, it sounds like?
Mayank Kapoor 32:42
Yes, not limited. It is important to us to always have very good relationships with all public cloud providers. If teams make the right case and justification for using a specific public cloud provider, they have that option. It’s not hard for them, we already have accounts there and it’s not hard for a team to choose AWS or Azure or GCP.
Reliance Jio: Cloud Consistency
John Allwright 33:11
So that’s actually a great segue into my next topic. When you’re operating on prem and in public cloud, even multiple public clouds, consistency must be key, and you said that Kubernetes is giving you that consistency as an infrastructure dial tone. Are there other elements of how you manage workloads in Kubernetes? How do you approach that kind of consistency, making sure that you have it across private-public cloud?
Mayank Kapoor 33:47
Yeah, so that has been a learning experience and a really important journey for us. We are very open source heavy. We try and coach the teams to not use too many managed services. In fact, some of our apps that are running in production are totally running on open source technologies, with no managed service. That makes transition between different clouds and migrations between cloud providers, which we do very regularly, much easier. Kubernetes is our computing layer now.
We haven’t started running databases on Kubernetes yet. Databases are still on virtual machines, just bare metal machines, we don’t use RDS for that, for example, a managed AWS relational database service. So that helps us keep consistency and move our workloads if we wanted to, to any public cloud. So that’s what is challenging, but that has helped us navigate the different cloud providers easily.
Jay Lyman 34:57
That matches the mix I talked about, where organizations are using cloud services, they are using some of their tools that they know and love from on prem and private clouds, and they are using open source. When we ask how they’re deploying Kubernetes, in addition to where, we see just that, it’s the public cloud distros, the open source distributions, and then the commercial distributions.
With that spread across there, there’s no question that Kubernetes is seen as a federating or a unifying control plane across these, and I don’t think we’ve ironed it all out. Kubernetes is very complex. You’re going to have to make a lot of configuration decisions as you deploy it, but that’s why we see reliance on the public cloud services and on the commercial distributions. That’s still playing out, but there’s still another example of a hybrid IT where it certainly isn’t a one-size-fits-all, and I think even within organizations, they’re running it across these different ways, whether it’s open source, public cloud services, or the commercial distros.
Reliance Jio: OpenStack, Kubernetes and VMs
John Allwright 36:17
So it’s interesting having Kubernetes as the universal substrate, but I’m sure there are some workloads that are happier on VMs. I’m interested, Jay and Mayank, what the role is of OpenStack? Obviously, that’s been providing private clouds for a while. Actually, Mayank, can you talk a little bit about your coexist with Kubernetes, OpenStack, how that works for you?
Mayank Kapoor 36:53
Sure. So, we have a significant OpenStack deployment, with 4000 bare metals nodes, 100,000 CPU cores, and growing. So yes, absolutely. Kubernetes is maybe 15-20% of our workloads. The rest is all virtual machines, and it depends on the team of course. Newer apps and newer teams will probably start working with Kubernetes straight away, but existing workloads and existing apps use virtual machines heavily for various reasons. Maybe they’re storing a state somewhere on that virtual machine. Maybe they need the local storage attached to the workload. They may need shared storage space.
There are lots of different use cases that still virtual machines just make it easy to solve. Especially snapshots. Snapshots are amazing with virtual machines, right? You can just take a snapshot of the virtual machine or your database, and then restore it if something goes wrong. Those kinds of flexibilities with virtual machines, it’s still there, and a lot of our teams use virtual machines still now. Virtual machines are not going away anytime soon.
Dev environments, especially, the developers love the virtual machines. We get one free virtual machine to every new developer that joins Jio, and that really helps. Especially with enterprise IT locking down our laptops a bit, at least they secure our laptops, so we don’t get to, you know, have full admin rights. So those kinds of things are really nice to have on the private cloud, where you have a virtual machine with full admin rights. So these kinds of things are very nice to have on OpenStack.
Jay Lyman 38:51
Yeah, I couldn’t agree more on the VMs. I think a lot of times when folks hear about uptake of cloud native and containers, they think that they’re displacing VMs. And my experience is that most containers are running on top of VMs or inside them, right, hopefully not in a one to one ratio. Oh, you will see that with large companies supposedly progressive. It’s a little bit like, you know, managing servers with spreadsheets. It’s something that we still see, but for the most part, organizations are finding lots of efficiency in slicing up VMs the same way, using containers, the same way that we sliced up physical servers with VMs.
I agree that VMs are still the currency of enterprise IT. There’s a lot of good tooling, security and capabilities around VMs. And by the way, they’re battle tested, right? Organizations know that they can trust those VMs.
The other thing to mention is that containers and VMs are influencing each other. We see lightweight versions of VMs that look like a container, or purpose specific VMs that are stripped down a little bit more like a container and more lightweight. So, I agree they’re evolving, and I often say, just as we have physical servers and even mainframes next to our VMs today, we will likely have VMs alongside our containers, you know, in the future. They’re not going away, they are getting sliced up, but what you’re saying I think makes total sense and you’re illustrating why the VMs have staying power. It’s some of that tooling, it’s familiarity, and it’s some of the things that containers can’t do as of this point.
Mayank Kapoor 40:47
One factor, I’d like to add, all of our Kubernetes clusters run on top of VMs. So, VMs are not going away, even if you’re doing Kubernetes. We have Kubernetes bare metal deployments, but the sheer luxury of just having cloud API’s to provision a Kubernetes cluster, destroy it, and provision it, again, is just phenomenal. So you can’t do it without virtual machines.
Running Kubernetes on Top of OpenStack
John Allwright 41:14
So, would you say on the whole, you’re running Kubernetes, on top of OpenStack?
Mayank Kapoor 41:21
We have some large, multi-tenant Kubernetes clusters running on bare metals that our central operations team manages. If you talk about self-serve, and you know, creating Kubernetes clusters on the fly, which a lot of our developers want, they want their own Kubernetes clusters with admin rights that they are running on top of VMs. We have automated the creation and destroy process for these clusters on OpenStack using cloud API’s.
John Allwright 41:52
I was reflecting, when you talk about developers liking the self service experience on prem, that’s probably coming from OpenStack. You may have abstracted that away slightly, but that’s kind of the stack that’s building up there.
You know, it’s interesting, the ratio between VMs and containers, it’s kind of like OpenStack and Kubernetes, are having a similar kind of dance. Like if folks know that a lot of customers are using both and you can obviously run Kubernetes clusters on a VM in OpenStack provisioned by OpenStack, but we also released a version of OpenStack, where the control plane runs on Kubernetes. So who is exactly on the bottom of the stack and who’s managing who? There are lots of options there, and it’s kind of a really interesting space, just to watch how you get the right infrastructure abstraction, and also the right cloud self service experience from the different technologies.
Jay Lyman 42:54
Yeah, a couple of years ago, there was a lot of talk about Kubernetes eclipsing OpenStack, wiping out OpenStack, whatever the case might be. What I thought was interesting was some of the most progressive and interesting Kubernetes use cases were running on OpenStack with Kubernetes. Whether it was a cloud provider, some large telecoms, obviously some service provider thing. I think, in attention, we can all agree that OpenStack was eclipsed by Kubernetes, but the two pair well, pretty good.
John Allwright 43:31
One of the perennial problems or challenges with OpenStack was upgrading. When it’s running, it’s fantastic. Upgrades are very complex and what does Kubernetes do really well? It lets you do upgrades with zero downtime. So by putting the OpenStack control plane on Kubernetes, it helped us remove one of the challenges with operating OpenStack. I see you nodding a lot Mayank.
Mayank Kapoor 43:57
Absolutely. I think the real challenge with running OpenStack was the day two and upgrades that you have to deploy. And yes, I think you guys are going in the right direction with running OpenStack on top of Kubernetes, just to solve that upgrade problem and make it just seamless and easy.
Choosing On Prem or Public Cloud
John Allwright 44:17
So just maybe one little discussion to segue us into Q&A.
What advice would you both have for folks that are considering their options with regard to on prem/public cloud? Maybe they’re still on prem, but haven’t gone cloud native, what should they think about when they’re considering that next move and their options?
Jay Lyman 44:47
You have to look at your new applications and the net new development and deployment, and your existing applications. If you’re a newer organization, like Mayank, you’re better off not being saddled with those, that older infrastructure that can lend you the workforce, and some of the skills and experience to make this journey so that you’re not looking to, you know, hire a Kubernetes army and become a Kubernetes company, but rather enable your teams with Kubernetes and with many different services that are out there. That means leveraging SaaS and the public cloud services, leveraging some of the tools that you’re used to, like monitoring or security that offers greater granularity or capability, or confidence. If you start with your net new, then you’re not so encumbered with the old way of doing things. Then, after you’ve mastered that a bit, you can go back and look, and that’s what we see.
One thing I wanted to mention with the VMs and the types of applications — we’re starting to see more stateful applications being containerized. When containers began, it was all about web and stateless applications, and I think those are still the bulk, but we are seeing more databases, more data servers, like Apache Hadoop and Spark and Kafka being containerized. That’s happened as organizations have applied cloud native to a wider swath of their applications and teams and as Kubernetes has evolved to support persistent data volumes, so it’s supply-side/demand-side is feeding this.
Starting smaller and with new, and then looking back and, you know, realizing that everything doesn’t have to go to the cloud, everything doesn’t have to be a microservice. There’s some applications that that’s not required for, and so it’s okay to run on VMs and to not necessarily use cloud native for all of your portfolio, but to instead search for that consistency, because that’s been a core part of our discussion here, that the more consistent you can make it across these different environments, the happier your developers will be, and the better off you’ll be.
Mayank Kapoor 47:34
Yep, I echo that as well. The advice I have for the central platforms organizations or the DevOps organizations within companies, would be instead of thinking either/or — either private cloud or public cloud — we should start thinking about multi-cloud. We should say it’s not either/or. I can plug into public clouds whenever I choose, and I can enable my teams, with the right security and compliance policies in place, on both private cloud and public clouds together.
So, the central DevOps team, or the platform team, has to build a platform that plugs into all these cloud services together. Then, Dev teams will choose which one they want and soon they will realize there are significant cost savings and justifications to have workloads running on private cloud, so they will gravitate to using private clouds as well. Even new teams. My suggestion would be for platforms teams to help enable developers with all the cloud services and cloud platforms they would want, and have the right networks in place, the pairings in place that allows them to do it.
That’s awesome. I mean, for the Platform VP whose CEO says, “We’re going all in on the cloud,” you have some very sage advice there. It’s not just about cost, it’s about experience. There are a lot of factors which are covered. So, great discussion. Thank you. I’m going to hand it back to Jay to take us through the Q&A.
Jay Lyman 49:26
How widely are telcos deploying Kubernetes for network functions like NFV?
Mayank Kapoor 49:56
Yeah, maybe I can take that, and John can help with other companies as well. So what I personally see, which may not be representative, is that the network companies are still figuring out container deployments for their network functions. They are in the early stages. We don’t have a network function, at least, to my knowledge, running on Kubernetes in production. There are definitely tests going on running network functions on containers moving into Kubernetes, but we don’t have one in production yet. So I would say it’s the early days and I think the network companies have to up their game and start moving a lot of their network functions to containers and Kubernetes just to stay competitive with normal application workloads. So that’s what I see in my company, at least.
John Allwright 50:59
Yeah, I’m not an NFV expert, although I do know it’s one of the areas that the telcos love OpenStack for because they can really build that cloud, that on prem cloud, and tune it to very high amounts of precision to suit their network requirements. At Mirantis, we do have a new release of OpenStack coming up in the next few weeks, with added NFV capability. So keep your eyes peeled for that, some exciting stuff coming there.
Jay Lyman 51:31
Does use of private cloud vary by vertical industry and/or geography?
John Allwright 51:46
So I think that might be one for the analyst.
Jay Lyman 51:52
We do see some industries more heavily involved in private cloud, and there’s a couple of drivers. You mentioned the telcos and one thing I always say is, you can’t really build your 5G network on Amazon, right? Then it wouldn’t be your network. So I think that’s a big driver. We also see organizations that necessarily don’t want to compete with the public cloud providers. That’s been a driver for private cloud.
Mayank, I’d be interested to get your perspective, you’ve talked about what drove your use of private cloud, but does anything line up with the vertical or the industry that you’re in?
Mayank Kapoor 52:41
Absolutely. So telcos, especially with the government regulations like in India, we have extreme data sovereignty requirements, so we cannot send our data outside India. That mandates our own clouds, our own data centers. We built our own network data centers across India, roughly 30 data centers across India. Anyways, we were good at running data centers. So for us, it was a natural step to say yeah, with our data requirements mandate, we put all data inside India, and none of the other public cloud providers are big enough to support our needs. So it was obvious for us.
I imagine most telcos are like that, most telcos will be using private clouds. For most startups, it wouldn’t be right. For most startups when they start out, the best choice for them is to go public cloud first, and then when they get big enough, they can consider a CapEx model rather than OpEx model.
So yeah, it’s definitely industry specific, the use of public clouds and private clouds. There’s a mix coming in where we see the other industries, where there’s not a strong case for either, so they’re evaluating both. If they have the CapEx funds, the CapEx budgets to invest in private cloud, I would guess they would do it because it makes sense.
John Allwright 54:15
Yeah, I’d echo that with some of the customers we see. With startups, it’s like the apps you start up, you start in the public cloud, because you don’t really know what capacity you need, you want to be able to flex. Once you know, particularly with the SaaS service, once you start to get traction, and if you’ve experienced hockey stick growth, then really the cost of the cloud is your raw material. It’s your cost of goods for the service that you’re selling. So, you want to try and minimize that as far as possible, and our experience is that the fast growing SaaS companies want to build their on prem capability, because they can really nail down that cost of the fuel, but they can also grow at the speed they need to grow as well.
Jay Lyman 55:00
What about Edge and IoT deployments? How are organizations viewing that type of venue and the infrastructure and applications required for it?
Mayank Kapoor 55:23
Very briefly, Edge and IoT for me, means low latency, high bandwidth, high throughput. That essentially is spelling out private cloud deployments for me. So if you need low latency, and your clouds closer to your users, it may be rare to have public cloud providers as close to the users as you want. So it’s very likely you don’t have public cloud providers in the same regions as your users. In that scenario, and especially for us, it makes a lot of sense to have our own private clouds running as close to our users for Edge and IoT workloads.
John Allwright 56:15
Another edge case, to use the word twice, another case that I’ve heard of recently is more industrial. You can think of wind turbines. Wind farms have a lot of processing going on locally, and maybe your chemical plant, where you actually want to put the processing quite near to the data so it can spin up some Kubernetes clusters there. They’re not next to users, but next to the data sources that need to be processed very quickly.
Jay Lyman 56:44
I certainly hear about more Edge and IoT use cases, particularly with cloud data. There’s a real match there with some of the capabilities of cloud native. I also see Edge and IoT has emerged as just yet another venue of hybrid cloud, right? We’ve been talking about on prem, talk about private clouds, public clouds, now we’re seeing Edge and IoT emerge as yet another venue that needs to be supported for many industries, more than just some of the industries that come to mind with Edge and IoT. We see a growing set of use cases there and so it seems to be emerging as yet another part of hybrid. So we’ll stay tuned to see where that goes.
Are enterprise organizations pulling back from public clouds in favor of moving back on premises or to private clouds?
Jay Lyman 57:54
We touched on this a little bit in our discussion, and I’ll just briefly say that we do see some of that. I think this is some of the public cloud fever breaking a little bit, and some of that bill shock that organizations get and realize that maybe we shouldn’t put everything in the public cloud, because we can’t afford it. But we do see some of that. Certainly not everything going to the public cloud is coming back, but I think, again, organizations have learned, they’re getting a little bit more pragmatic and decisive about which workloads should run where.
Mayank Kapoor 58:35
Yeah, I definitely see workloads moving back to the private cloud, especially now that Kubernetes has been a game changer. So, the same experience that teams get on the public cloud, they’re also getting on the private cloud. So that I definitely see.
John Allwright 58:57
I think as time goes on, we’ll see more examples of what can happen with public cloud, whether that’s outages, or the control that the public cloud has over the workloads running. And then, folks in retail, maybe don’t want to host on a particular public cloud, because it’s competition, but there are more examples of that. People are getting much better about really feeling not just intellectually, but feeling the risk that they might be exposing themselves to on public cloud. There’s the rewards as well, but actually understanding it in a much more tangible way.
Jay Lyman 59:37
Good. Well, that’s about all we have time for. We’re going to have to wrap up the discussion there. We will answer your questions that we weren’t able to get to via email, but we thank you, John, we thank you, Mayank, for joining us. Thanks to our audience for your time and interest. Be well and have a good day.