 Welcome back everyone to SuperCloud 22 live in the Palo Alto office our stage performance. We're streaming virtually It's our pilot event our inaugural event SuperCloud 22. I'm John Furrier with my ghost Dave Vellante I got a featured keynote conversation with Kit Colbert who's the CTO of EM where to lay it all out break it down kid Great to see you. Thanks for joining us for SuperCloud 20 our inaugural event. Yeah, I'm excited to be here Thanks for having me. So we had great distinguished panels coming up through we heard Victoria earlier to the keynote There's a shift happening the shift has happened. That's called cloud You just published a white paper that kind of brings out these new challenges around the complexity of how companies want to run their business Yep, it's not born in the cloud. It's cloud everywhere seems to be the theme. What's your take on SuperCloud? What's the roadmap for multi-cloud? Yeah Well, you know the reason that we got interested in this was just talking to our customers And the reality is everybody is using multiple clouds today multiple public clouds. They got things on-prem They got stuff at the edge and so their applications are essentially distributed everywhere and The challenges they start running into there is that there's just a lot of heterogeneity there It's like different API's different capabilities inconsistencies incompatibilities in terms of workload placement data migration security as we just heard about etc And so I think everyone's struggling with trying to figure out How do I drive consistency across all that diversity and what sort of consistency do I want? and one of the things that became really interesting in our conversations with customers is that There is no one-size-fits-all that different folks are in different places and the types of Consistency that they want to prioritize will be different based on their individual business requirements And so they started forming a picture for us saying okay, what we need are a set of capabilities of you know multi-cloud cross-cloud services that deliver that consistency across all the different environments where their applications may be running and That is what formed that the early thinking and Sort of the paper that we wrote on it as well as some of the work and that I think eventually Leads to this vision of super cloud, right? I think you guys have the right idea, which is Hey, how does all this stuff come together and and what does that bigger picture look like? And so I think between the sort of the native services that are there individually for each cloud that offer great value By the way, and people definitely should be taking advantage of in addition to another set of services Which are multi-cloud that go across clouds and provide that consistency looking at that together That's in my picture where super cloud is So the paper is called the era of multi-cloud services arrived VMware executive outlook for IT leaders and decision-makers I'm sure you can get on your website. Yep And in there you talked about well first of all I think you would agree that multi-cloud has fundamentally been a system a symptom of Multivendor or M&A. I mean you talked about that in the paper, right? Yeah, it was never really a strategy It was just like hey, we woke up in the 2020s and here we are with multiple clouds, right? Yeah, it was one of those situations where most folks that we talked to didn't plan to be multi-cloud Now that's changed a little bit in the past year or two sure But certainly in the earlier days of cloud people would go all in and say hey I'm gonna go all in on one of the major hyperscalers and and go for it there And that's great and offers a lot of advantages, right? There is internal consistency there There's usually pretty good integration between their services so on and so forth The problem though that you start facing is that to your point Acquisitions you acquire companies using a different cloud. Okay. Now I got two different clouds Or sometimes you have the phenomenon of shadow it still happening where some random line of business is going to go off And use a different cloud for whatever reason The other thing that we've seen is that over time that you may have standardized on one But then over time technology changes another cloud makes major advancements In the state of the art or you know, let's say in machine learning and you say hey I want to go to the other cloud for that so what we start to see is that people now are choosing public clouds Based on best of breed service capabilities and that they're gonna make those decisions and fairly fine-grained manner Right sometimes down to the the team the line of business, etc And so this is where customers, you know companies find themselves now It's like oh boy now have all these clouds and what's happened is that they kind of dealt with it in ad hoc manner They would spin up Individual operations teams Security teams etc that specialized in each of the clouds they had knowledge about how to do that But now people found that okay. I'm duplicating all this There's not really consistency in my approach here. Is there a better way? And I think this is again the advent of a lot of the thinking of multi-cloud services and super cloud And I think one of the things too and listening to you talk is that the old model used to be you know Solve complexity with more complexity Okay, and customers don't want that from what we're observing and what you're saying is they've seen the benefits of dev ops dev sec ops So they know the value yeah, because they've been in say one native cloud now They say okay I'm on premise and we heard from Vitoria said there's a lot of private cloud going on But since he makes that another cloud out by default as well So hybrid is multi-cloud hybrid as a subset. Yeah we kind of have this evolution of thinking right where you kind of had all this sort of Different locations and then we I think hybrid was an attempt to say okay Let's try to connect one location or a set of locations on premises with a public cloud And have some level of consistency there But really what we look at here with multi-cloud or super cloud is that that's really a generalization of that they're not talking about one or two locations on prem and one cloud we're talking about everything now and Moreover, I think hybrid cloud tended to focus a lot on sort of core infrastructure and management This looks across the board. We're talking about security. We're talking about application development talking about end user experience things like zero trust We're talking about infrastructure data. So, you know, it goes much much broader I think then when we talked about hybrid cloud a few years ago So in your paper, you've essentially kit laid out an early framework. Let's call it for For what we call super cloud what you call we go cross cloud services So what what do you see as the technical enablers that are you know the salient aspects of again? Multi-cloud or super cloud. Yep well, so for me it comes down to So okay taking a step back So we have this problem right where you have a lot of diversity across different clouds and customers are looking for some levels of Consistency, but as I said Rarely do I see two customers that want exactly the same types of consistency and so what we try to do is step back and first of all Establish a taxonomy and that by that by that I mean one of the different types of consistency that you might want and so There's things around infrastructure Consistency security consistency software supply chain securities that probably the top of mind one that I hear from customers application and application services so things like databases Messaging streaming services a iml services, etc And user capabilities and then of course data as well And so you know in the paper we say okay Here's these kind of five areas of consistency and that's the first piece the second one then turns more to an architectural question of What exactly is a multi-cloud service? What does that mean for a cloud service to be multi-cloud and what are the properties there? So essentially said okay, we see three different types of those There's one where that service could run on a single cloud but could support multiple clouds So think about for instance a service that does cost analysis now it may have you know, maybe executing on AWS Let's say but it can do cost analysis for Azure or Google or AWS or anybody right so that's the first type The second type Is a bit more advanced where now you're saying I can actually instantiate that same service into multiple clouds And we see that oftentimes with things like databases that have a lot of performance latency, etc Requirements and that you can't be accessing that database remotely that doesn't you know from a different cloud That's going to be too slow you want to be in it have it on the same cloud that you're in and so again You see various vendors out there implementing that where that database can be instantiated wherever you like and then the third one Would be going even further and this is where we really get into some of the more much more difficult use cases where customers want a workload to be on-prem and sometimes especially for those that are You know very regulatory compliant. They may need even in an air-gapped or disconnected environment So there can you take that same service, but now run it without your operators being able to manage it 24-7 So those are the three categories sort of a single cloud supporting single cloud instance supporting multiple clouds Multi-cloud instance multi-cloud instance disconnected so you're abstracting you as the you know the R&D arm You're abstracting that complexity. How do you handle this problem where you've got one cloud Maybe has a better service than the other cause do you have to devolve to the lowest combat denominator or yeah? Or how do you mask that? Well, we so that's a really good question And we debated it and there's been a lot of thought on it Our current point of view is that we really want to leave it up to the the company themselves to make that decision Again because we see different use cases So for instance I talked to customers in the defense sector and they are like hey You know if a foreign adversary is attacking one of these public clouds that we're in we got to be able to Evacuate our applications from there sometimes in minutes right in order to maintain our operational capabilities And so there there doesn't need to be a least combat denominator approach just because of that requirement I see other folks, you know, you look at the financial banking industries. They're also regular I think for them it's oftentimes 90 days to get out of a class So they can do a little bit of rearchitecture you got times to roll the sleeves and change some things So maybe it's not quite a strict whereas other companies say you know what I want to take advantage of these best of breed services native to the clouds So we don't Try to prescribe a certain approach there, but we say you got to align it with what your business requirements are How about the PAS layer? So one of the things we've said is that we felt like a Super PAS was a was a requirement of Super cloud because it's a it's a purpose-built PAS that Helps you with that objective whatever that is and you say in the paper for developers each cloud provider has unique infrastructure Interfaces and API's that add work and slow the pace of their releases for operators each additional cloud increases the complexity of their architecture Fragmenting security performance optimization and cost management. So so are you building a super PAS? What's your philosophy? You know Vittorio said have we want to have our cake. We want to eat it, too And we want to lose weight. So how do you do that? Yeah? So I think it's so first things first, you know What the paper is trying to present in the end is really sort of an architectural point of view on how to approach this right and then yeah We at VM where we've got a lot of solutions towards some of those things But we also realize we can't do everything ourselves right the space is too large So it's very much a partner strategy there Now that being said on things like on the PAS side, you know We are doing a lot for instance around Tanzu which is our modern apps portfolio products and the focus there really is to yes Provide some of that consistency across different clouds Enabling customers to take advantage of either cross-cloud PAS type services or Cloud native or native cloud services, I should say And so we really give customers that choice and I think that's for us where it's at because again We don't see it as a one-size-fits-all. There's your cake and eat it, too So you're saying the developer experience can be identical across clouds unless the developers don't want it to be Yeah, and maybe the team makes that decision. Look, there's a lot of reasons why you may want to make that or may not The reality is that these native cloud services do add a lot of value and oftentimes are very easy to consume to get started With to get going and so, you know, it's a trade-off. You got to think about and I don't think there's a right answer It's okay I got to ask on you you said you can't do it alone. Yeah and where I know for a fact You guys have been working on this for many many years. Yeah, Ragu remember I interviewed him in 2016 when he did the deal With AWS with Andy Jassy that really moved the needle things got really great from there with VMware So would you would you be open to a consortium to? Overseek is you guys a lot of a lot of investment in this as a company But I also don't hear you trying to do the lock-in thing So yeah, would you guys be open to a consortium to kind of try to figure out what? These buildings blocks look like or is it a bag of Legos is what people want absolutely and you know what we Offer in the papers really just a starting point. It's pretty simple. We're trying to define a few basic items of the taxonomy and some Outlines sketches if you will of what that architectural picture might look like but it's very much that like just a starting point and This is not something we can do alone This is something that we really need the entire industry to rally around because again I think what's important here are standards. Yeah that that there's got to be This sort of decomposition of functionality break down in the different sort of logical layers of functionality What are those apis or interfaces look like? How do we ensure interoperability because we do want people to be able to get the best of breed to be able to bring together Different vendor solutions to enable that and I was watching and it was had a silicon A day just last week talking about their advances and silicon What's your guys position on that because you're seeing the I as players almost getting more nichey and more better at Hardware matters more silicon speed latency GPU. Yeah, so that seems to me be an enabler opportunity for the ecosystem To innovate at the past and sass relationship What where do you guys see? Where are you guys? Strong and where do you need work to do on if you had to say there was some white space sure that vm Where let me say hey we lot we own this area. We were solid here Here's some white spaces that vm where could use some help with yeah Well, I think the the infrastructure space you just mentioned is clearly one that that we've been focused on for a long time We're expanding into the modern app space expanding into security. We've been strong and end user for a while So a lot of the different Multi-cloud capabilities we've actually been to your point developing for a while And I think that's exactly again what went into this like what we started noticing was all of our different product teams Were reacting to the same thing and we weren't necessarily talking about it together yet They're like what well this this whole challenge of multiple clouds of dealing with that heterogeneity of wanting choice and flexibility into where to place a workload or where to place a virtual desktop or you know Whatever it might be And so each of the teams was responding individually To that customer feedback and so I think what what we recognized was like hey Let's let's up level this and what's the bigger picture and what's the cut sort of common architecture across all of it, right? So I think that that's what the really interesting aspect here was is that this is very much driven by what we're hearing directly from customers You kind of implied just recently that you know the the paper was pretty straightforward a pretty basic early days But but it's well thought out and one of the things you talked about was the type of Multi-cloud services your data plane and user services security infrastructure, which is your wheelhouse and application services So and you sort of def wanted to you know detail defining those Where is management and all that yeah, these are the ones you're going after what about management? What are your thoughts on that? Yeah, so it's a really good question. We debated this for a long time Does management actually get a separate? Sort of layer that you could we could add a six one perhaps or is it sort of baked in to the different ones? And we kind of went with the ladder where it's sort of baked in there's infrastructure management There's modern app management There's management and users kind of management, you know for each security obviously So we see a lot of different management planes control planes across each of those different layers Now doesn't need to be a separate one That has its own layer arguably yes I mean I think there are good arguments for that and this is exactly why we put this out there though It's like get people to read it keep able to give give us feedback and going back to the consortium idea Let's come together as a group of practitioners across the industry to really figure out an industry viewpoint on this So what's are the trade-offs there? So there's what would be the benefit of having that separate layer? What I presume it's simpler to do it the way you've done it But what would be the better of having a separate I think it was probably more about simplicity to start with like you Couldn't imagine like 20 different layers and and maybe that's you know where it's gonna go But also I think it's Yeah, I think it's how do you define the layer and for us It was more around sort of some of these functional aspects as an infrastructure versus application level versus end user And management is more of a common ally across those but again, I I could see our logical place to start Yeah, the other thing you said in here multi-cloud application services can route request for a particular service such as a database and deploy the service on The correct individual cloud using the most appropriate technology for the use case etc. Etc. That to me sounds like a metadata Problem and and so you can you talk about how you you've approached that you mentioned AWS RDS Great examples as your sequel or an oracle database etc etc multiple endpoints. How do you approach that? Yeah? Well, I think there's a bunch of different approaches there and so again So the idea is that and you know, I know there's been references sort of like the operating system for super cloud What does that look like right? But I think it totally you know, we don't actually use that term But I do like the the concept of an operating system because a lot of things you just talk about there These are things operating systems do you got to have a scheduler and so you look across Many different clouds and you got to figure out, okay Where do I actually want in this case? Let's say a database and since to go and be provisioned and then really it's up to I think the vendor or the In this case the multi-cloud service creator to define how they want to do that They could leverage the native cloud services or they could build their own technology, right? Which a lot of the vendors are doing and so the point though is that really you get this night from a end user standpoint It goes back to your complexity simplicity question. You get the simplicity of a single API That the implementation you don't really need to deal with because you're like I'm getting a service And I need the database and has certain properties and I want to hear versus there versus wherever But it's up to that multi-cloud service to figure out a lot of those Implementation specifics. So are you the super cloud OS? I think it is VMware's goal to become the super cloud OS for sure, but you know like any good operating system as we said like It's all about applications, right? So you have you have that you have a platform point of view, but you got a partner widely and and you got to get the hardware Relationship yes, the silicon chips. Yep, right. Yeah, that's a good point I want to go back to that one if you mentioned that earlier the innovation that we're seeing, you know things like Arm processors and like Graviton and a lot of these things happening And so I think that's another really interesting area where you're seeing tremendous innovation there in the public cloud One of the challenges though for public cloud is actually at scale and then it takes longer to release newer hardware at that scale So in some cases if you want bleeding edge stuff, you can't go with public cloud because it's just not there yet, right? So that's again another interesting thing where you well Some will say that they want 5,000 new services every year out of the adbs. No, but I have some bleeding edge stuff I'm not talking about the software. So I'm talking about the hard work. Okay. Okay. So like the silicon Yeah, like, you know the latest and greatest GPU FPGA Why why can't they? Because they do like tens of thousands of hundreds of thousands of them. Oh just because it's a scale. Yeah, that's the point Right, and it's it's fundamental the model right in terms of how big they are And so that's why we do see some customers who need who have very specialized hardware requirements Needing to do it in the private cloud right on prem or possibly to Colo or edge or edge is a great example Of but we often see again people like, you know, the latest bleeding edge GPU is whatever they are Even something a bit more experimental that they're gonna go on on prem for that And so look, we're not I'm not do not want to disparage the public clouds. We don't take that away It's just an artifact when it gets so hard like software they can scale and they do well It's in context of the OS conversation OS has to write to hardware and enable applications where I was getting caught up in that is kid is that they're all developing their own Silicon and they're developing in a place But most of its arm-based and they develop in a much much faster cycle They can go from design to tape out much faster than Intel historically has and you're seeing it You know Intel just posted a lot. I think if you look at the overall system, you're absolutely right Yeah, but it's the it's the deployment because of the scale because that one Availability zone and another and another region and that's yeah So a counterpoint to what I just said would be hey like they have very well-controlled environments very well-controlled systems So they don't need to support a million different Configuration settings or whatever they've got theirs they use right so from a system standpoint and so forth Yeah, I agree that there's a lot they can do there I was speaking specifically to you know different types of hardware accelerators Being a bit of a bottom if it's not in the 5,000 services that they offer you can't get it Whereas on-prem you can say I want that yeah, here it is I'm not saying that on-prem necessarily fundamentally better in any way I'm just saying for this particular area use case-driven it is you and that's the whole point of all this right like And I know you know a lot of people in their heads associate VMware with on-prem But we are not dogmatic at all and you know as you guys know but many people may not like we partner with all the public cloud hyperscalers and So our point of view is very much much more nuanced saying look we're happy to run workloads wherever you want to in fact That's what we hear from customers. They want to run them everywhere But it's about finding the right tool for the right job, and I think that's what really what this multi-cloud approach Yeah, and I think the structural change of the virtualization hypervisor this new shift to v2 super cloud This something happening fundamentally. Yeah, that's use case-driven. It's not about dogma and whatever I mean clouds great native clouds have the pros and cons I would say that super cloud Pre-requisite for super clouds got to be running in a public cloud But I'd say it also has to be inclusive of on-prem data And you're not going to just move all that data into the maybe in the fullness of time But I don't personally believe that but you look at what Goldman Sachs has done with AWS They've got their on-prem data, and they're connecting to the AWS cloud Yep What what Walmart's doing with Azure and that's going to happen in a lot of different industries Yeah, well, I think security will drive that too. We had that conversation because no one wants to increase the surface area Number one they want complexity to be reduced and they want economic benefits. Yeah, that's the super cloud It's a security, but it's also differentiable advantage that you actually have on-prem that you don't all right Well, we're going to debate this now kid. Thank you for coming on and giving that keynote We're gonna have a panel to debate and discuss the blockers and enablers to super cloud And and there are some enablers and yeah, thanks. We blockers. Yeah, so we'll get into that Okay, up next the panel to discuss blockers and enablers of super cloud after this quick break