Bogomil Balkansky, a Google engineer who came from Bulgaria to VMware, traces cloud computing from virtualization (Cloud 1.0) through containers (Cloud 2.0). He argues that true cloud transformation requires compute, networking, and storage to become software-defined, but data is the hardest piece because of its "gravitational pull" and the difficulty of moving massive data stores mid-flight.
Chapters
Let's talk a little bit about yourself, maybe tell us a little bit about your background.
Well, Mark, like you and so many other people in the Valley, I'm also an Emmy Gray and a first generation American. I grew up in Bulgaria and came here for college and kind of never left.
So you came here and pretty soon you got involved in one of the biggest revolutions in the enterprise in about a decade. You joined VMware. That must have been an exciting time.
Yeah, I absolutely consider myself lucky that I got to participate in that first wave, I would call it, of cloud infrastructure that VMware created in the 2000s. I would say that this groundswell that we're seeing today, the transition towards cloud anything, be it software as a service, applications, be it cloud infrastructures. Really the genesis of that was the invention and the commercialization of the virtual machine by VMware. This was like the first instantiation of showing how it provisioning and management can be done a lot more easily, a lot more faster, greater agility if you transform hardware into software.
So you've really been in the center of what perhaps you might call now Cloud 1.0. Data centers became software defined, so to speak. And it completely changed the paradigm.
Yes, it did completely change the paradigm. Now the interesting thing is that change didn't occur either quickly or easily.
And now we're seeing what the next tsunami, so to speak, Cloud 2.0, occasionally it's being called. It's all kind of based on containers and they're really penetrating, proliferating in the enterprise, in the infrastructure, at almost unprecedented speed. What's driving this? What is the challenge that people are trying to solve?
I would say that the transition towards cloud in general really comes down to two things. One is simplification that allows you to do things a lot faster, or people usually call that agility. And the second thing, it's about a new business model which is about pay as you go, pay for only what you consume. And this is a really drastic change in terms of how people think about the economics of infrastructure. But the speed with which containers have really emerged as the new compute unit, I think is yet another proof point in that trend, or another step in that trend of simplifying things. The VM is stateful, it's a lot heavier duty container than the container. And so it's really kind of the next logical step in terms of making the infrastructure lighter weight, easier to provision, stateless, etc.
Just a compute orchestration is not a complete solution. There are other pieces required to really fulfill that dream. To fulfill that vision.
Yeah, absolutely. You're right that the revolution that VMware started with the VM was obviously on the compute side. But the transition to cloud infrastructures, it's not complete until the triad of compute, network and storage data is also transformed to behave in this new agile and dynamic way. When it comes to this transition, compute all these leads, then networking and security functions, function follow after that. And I would say data is the last one to the party, so to speak, for a good reason, because as you mentioned, it has the most gravitational pull. It is really the crown jewel of any company's assets. Right? It really, in this day and age where everything is digital, the data lake, or whatever you want to call it like, of any company, really represents anything and everything that the company is about. And that assets need to be carefully protected and handled very carefully. And because companies just have tons and tons of data, it is very difficult to be evolving that infrastructure. Just imagine you're flying Boeing 747 and as you're flying it, you're trying to make changes to it. You can't swap out the engines in mid flight. And obviously this is a folksy sort of metaphor, but I think it really illustrates the challenge of transforming the data. I think it is absolutely the hardest part in order to achieve that complete cloud transformation.
And let's talk about that range of applications that we're seeing in the enterprise. We still have lots of Cloud 1.0, VMware. We're seeing increasingly Cloud 2.0, Cloud container based. The diversity of applications, the spectrum of applications, the spectrum of platforms, the spectrum of orchestration frameworks is only increasing, the complexity is only increasing. What does that mean for your infrastructure, specifically the data infrastructure underneath, to be able to handle that, to be able to serve all of that.
Now it's interesting that in it, nothing ever disappears. Whatever you bring in the data center, it just tends to stay there forever. Which basically means that the heterogeneity and the diversity of the kinds of applications and workloads that companies need to deal with just increases over time. But because this heterogeneity of application and workloads is increasing, that really makes it very, very difficult to transform the data infrastructure. Because the implication is that the data infrastructure needs to be able to respond to all sorts of different service level requirements from different applications. Let's not forget that the service level requirement for any application is not something that's fixed over time. It's very dynamic. It changes by the hour, by the month, et cetera. So that means that the data infrastructure needs to become a lot more intelligent to be able to very quickly understand and dynamically provision to the SLAs that are required by this increasingly heterogeneous corpus of applications. That also means that data in essence needs to be provided as a service. Just like compute is provided as a service through containers. Just like networking increasingly is provided as a service through all the software defined networking innovations, Data also needs to get there in order to complete that transition to cloud infrastructure. Sometimes people talk about the elasticity of cloud infrastructure, so the data infrastructure needs to be able to respond both to the SLAs as well as to the throughput the scale requirements that different workloads have.
Bogomil, thank you so much. It's been such a pleasure to have you here. Thank you, thank you.
