VMware Cloud CTO Marc Fleischmann on Data Management for Multicloud

foreign speaks we're talking about data management in a multi-cloud world data management has always had a long list of challenges getting proper Access Data quality governance and compliance but in today's multi-cloud world with data distributed across a mixed environment these challenges are getting even tougher to discuss that I'm joined by a very special guest with me is Mark Fleischmann CTO for VMware Cloud Mark actually thrilled to have you with us today thank you for having me James really looking forward I think that companies realize these days they're getting the most from their data is definitely very very important that they need that for competitive Advantage but the problem is is that some of the strategies created years ago aren't quite keeping up with today's you know mixed multi-cloud world why is that so why is today's multi-cloud world offering some new challenges for data management yeah this is a multi-layer question and so let's sort of take it in stride a little bit James well first of all uh maybe just to give some perspective data really is at the core of everything it's at the core of digital transformation right the ability to extract value from data is really more and more determining the at the success of modern Enterprises so that's what makes them think about what they did a managed management strategy needs are and what they should like look like in this age of of of multi-cloud computing what's one interaction Mark before you go on I think an important point about you say data is everything data is really the foundation of artificial intelligence you know I hear all this Buzz about ai ai goes nowhere where that an enormous data repository feeding it so I totally agree data is really everything these days absolutely and we should talk a little bit about AI as well but even before you get to that um and before you get to multi-cloud even the way data is consumed today not only by AI but by uh um you know modern applications it's very different it's changing so we've kind of seen an evolution from traditional data storage that's the block and file traditional sort of you know arrays right we have a product there too vsan which are quite innovating a little bit you know to object and then mostly we're sort of getting to information which is sorting the data and knowledge which is classifying the data and then actually running AI on it right and especially classifying the data is really uh the domain of of databases um and so Cloud infrastructure is really expanding Beyond just storing data to sort of managing sharing and analyzing it and again AI is is obviously uh part of that so what why why is multi-cloud itself which seems to you know Federate the data distribute across why is that a new challenge in today's world so um as modern applications are sort of uh raising the abstraction to to databases and you get to data infrastructure um that sort of operationalizes the databases and by the way VMware is is doing a lot there um VMware Explorer is coming up our our annual um uh conference we're gonna make some pretty big announcements in in data management uh how data storage manager um we're gonna make some pretty data services manager we're going to make some pretty big announcements there about how we manage rapidly growing growing DB Estates and sort of how we allow app teams to self-service these DB States a little more and and we should talk about that a little too the next wave after data infrastructure is really data management and what that means is you now take these these databases these data States and you start connecting and integrating them across the cloud as silos if they post silos by by themselves and as you integrate and connect this data you really take it from information and knowledge to actionable insights and this is why this is such a big deal and such a such a kind of Paradigm Shift there right well I think I mean you use the problematic word which is silos I think that's the thing that the companies are always afraid of like oh it's they're in silos in this Ideal World they might physically exist in silos but they're not actually silent in terms of being able to access the entire repository of data so it is silo not the enemy of data data mining at some level oh absolutely I think it's one of the key challenges that we really uh need to to uh attack here right and siloing doesn't mean only physical siloing but I mean by the way it could mean operational siloing it could mean security siloing it could mean um um uh it could be tools that that are siled um compliance silos um so sort of getting beyond that is really a significant uh engineering effort and something that's uh really very very important as I said extract the value value of this data more broadly I think it's also something that is right in the Wheelhouse uh for VMware because you know think about what we do we you know we manage infrastructure and that can include data infrastructure as well be lifecycle management we update it we provide Telemetry and monitoring on it we do remediation we do backups so all the stuff that we do for for infrastructure by itself we can increasingly apply to data infrastructure running on top of it which by the way becomes it's becoming a significant part of we have the Telemetry to see this it's becoming a significant part um of the applications we run and as we apply it to data infrastructure running on top of it then you know you start connecting it with messaging streaming and as I said that's when you get to data to this integration and sort of data connectivity connected data well okay then I I almost feel bad asking this question because I think the answer probably would take about four four hours to really give a full answer to but companies face a problem with data management in this distributed environment any tips or best practices for for improving their data management in in this mixed world and again I know this is It's a vast question but is it what can you say in a nutshell to that question yeah it's a it's a vast question um I think um Let Me Maybe focus on one or two aspects of it um one that's pretty important and I mentioned it briefly is that data security um we see this you know talk about llm's large language models um in a new manifestation where sort of you know some of your you know assets some of your secret sauce might not leak out into into these into these models so having a comprehensive security strategy across all these Cloud silos and managing it end to end perhaps policy based and forcing those policies making sure that only what you want actually ultimately um it becomes visible becomes accessible um becomes accessible to the right audiences they vary as well I think that's a very very big aspect and you need that um uh coherent across all these all these silos so I think I think that's that's one big aspect well you know data security sometimes I hear a lot about there's a conflict going on there where the data security is on one side data governance and data security on the other side is getting fast and easy access to all throughout the organization maybe it was a middle manager that needs Fast Access so and of course people need to access the data faster and faster than ever before so data security versus data access do you see that conflict going on yeah absolutely and honestly this that's a a topic I personally like a lot because what that gets you to is orchestrating um ideally policy based um coherent policy across all those pools and if you have a coherent unified infrastructure like you know our multi-cloud architecture our multi-cloud platform on the VMware Cloud platform if you have a coherent infrastructure across it and you can drive coherent policy across it and then obviously you have to you have to your consumers and your producers defined you actually can at least help addressing this challenge between data security and data access in a pretty comprehensive and Powerful way uh something that I'm personally focused on a lot and and it's a it's something I can get quite passionate about you know I I don't know my what my question is there is is part of the question is it giving it individuals certain credentials is it building those credentials into the system how is this is there a technical answer to the question how do you ease that conflict between data security and data access um but the the way you you do it is obviously it does come down to individuals but individuals are part of organizations and or groups and those groups are parts of organizations and so um you can sort of take this role-based access airbag right our back and you can expand it into a hierarchical sort of model um how you um gradually refine access rights for organizations as you sort of map them into into this hierarchy um and you give them guardrails and governance rules and you could get very very sophisticated there and again this is sort of the kind of system we're building there so that um operators basically ultimately have the ability from a single pane of class to Define this access across multiple clouds um coherently for their organizations down to single individuals if you if you want so uh and again it's it's very very powerful well let's drill down to the VMware product offering and you've mentioned a VMware how is VMware addressing the cloud-based data management needs of its clients what what's the the VMware advantage well again um I would say data access really goes uh in in multiple ways right you look at at storage um what we're doing there is we're innovating on the storage layer by itself we're taking storage from like I would say on-prem or just cloud-based assets to more hybrid starch assets that can actually span where you have data planes that span multiple locations um that's a quite some engineering there you'll also see some announcements here we've launched what we call a vsan Esa our Express storage architecture which really takes storage to the next level it expands it way into sort of the mid-range capabilities this gives one already unique um advantage to VMware even at the the start at the foundation of everything data storage because we now can span traditional applications with an Enterprise truly Enterprise class uh data plane but then at the same time gets is consumable um through cloud-like uh through a cloud-like experience through cloud-like operations right so it really sort of spans the whole gamut between traditional storage and cloud-like storage so that that's number one right then as we uh go up to the next level and we get to data services we manage the data services on our platform we start with Queen Plum and rabbit and Q green plum is the data their house rabbitmq is our um is our messaging service and we're actually adding um more databases um that are are really very popular out there that includes postgres MySQL mongodb um and we'll also add object storage right and we'll manage them with the richness of the the VMware management capabilities including monitoring Telemetry as I mentioned before life cycle management remediation so that gets interested data services which now uh give modern applications the consumption model that they like and the last step then is um data data management um itself where you then start integrating these data sources and and that really gets us into into message cues into um into into queuing into into streaming capabilities that that we that we can put on this platform as well across spanning these Cloud silos interesting you know thinking about accessing data in various different environments I I think about the fact that I've covered uh it since even before the world of cloud believe it or not I remember those VMware in in the pre-cloud days and there was the kind of the excitement of the fact that what VMware was doing which is it was extra it was abstracting the hardware from the software and that was at the time it was so revolutionary and really changed everything I guess I think about managing data these worlds these days in this world of of mixed environments maybe the point is to abstract the various environments so that the user or the administrator has one panel that the data could reside at any different any different number of places but it's it's abstracted does that make any sense James I I love that analogy and if I may I I use that at VMware myself so I I keep talking about I would say maybe the First Act of VMware was sort of maybe call it vertical virtualization where you take a machine and you obviously sort of fragmented and sort of basically within still these Hardware the confinements of these Hardware boundaries you you sort of uh you virtualized uh the machine for the workloads but now with Cloud this whole thing sort of flipped and we're now talking about horizontal but I would call horizontal virtualization it's aggregation and then sort of pooling of these resources and farming them back out and you're right it's another in mind I completely agree it's another form of virtualization that's really the bottom line it's obviously a lot more management and specifically Orchestra Asian intense because now you need to sort of orchestrate these pretty massive scaled pools and sort of deliver them uh judiciously you know at scale to your much larger user base as well but yeah it's a it's in my it's I very much agree it's another form of virtualization interesting I have never thought of it that way there there was the classic original like early 2000s VMware virtualization you might think of it as vertical virtualization because there's a server then that sits right on top of that server immediately but you're using the freight horizontal virtualization to refer to virtualizing a really mixed you know even Geographic across the cloud world yes yes uh again that that's how I refer uh to at VMware and that's sort of how to me why this is such a so so much in the Wheelhouse of what VMware does because we do virtualization via the virtualization company right um and yeah we started vertical and the world evolved and you know we compute gets proliferates everywhere it becomes more accessible just as you said before we we all started earlier at some point this was called a utility I I think we now have the compute utility it's called the cloud used to be utility in the 90s right but that's again all about aggregation and orchestration and and ultimately that's another a form of virtualization and managing those virtualized resources well all right on those lofty thoughts let's look at the future instead of a utopian Vision if we think about data management and multi-cloud a few years out maybe two to three years out or further depending on how Brave you want to be what what do you see when you see the future of data management in a mixed environment but I really do see um it's sort of expanding Into You know here we go again ml Ops data Ops data meshes um and data data grids right um I see it in a way these are there's some hard challenges by the way right because the more heterogeneous your data sources that you need to read from and and your data destinations that you need to write from uh become um the the more sophisticated the software you need to put on top of it becomes right so just to give you a few examples um uh reading data has challenges like handling multiple query dialects and protocols maintaining data catalogs across these DB schemas keeping up with the revolving DB schemas they're basically sort of you know they're they're obsolete almost when they're created managing resource contention across those silos a very hard problem because they're so heterogeneous um assuring data recent see assuring data compliance data quality it goes on and on okay writing data also has a number of challenges like solving transaction management and so the business logic transformation so these are very tough problems that we are working hard to address at the next level and this is really forward looking so I can't go too much into it James that way you think like well you know maybe there's ways to think about abstractions that can sort of span these data pools and sort of hide the complexity underneath you can think almost again as the next level virtualization what's the data model that can sort of span all this and sort of abstract all it and make it and make it very broadly and very uniformly accessible and I think that would be very powerful well I hadn't I did mesh of course it seems like it's a very trendy term it's on the way up and of course data mesh is not exactly a technology it's an architecture for for data infrastructure I it never occurred to me as you mentioned that the software needs to be more advanced to handle that data mesh world um that's an interesting thought the other part is it seems like we're going to ask more of artificial intelligence in terms of AI is going to be doing more and more of the work of handling the distributed data any thoughts on that yeah yeah look it sort of fits into I would say our longer term vision is to make VMware the de facto multi-cloud data management leader or you know maybe even set standards there with the data management platform that helps consumers locate access utilize DB's data across data warehouses and data lakes and various data sources right um and so now we get to AIML which as you pointed out at the beginning of our conversation at its foundation is is a data management or a data problem right the the more data you have to feed your AIML models or ml models specifically or llm models for that for that reason right um for that matter uh the the smarter they become and and so you get back to like how can you aggregate how can you best aggregate all this data from these various pools and sort of curate it and and prep it I mean that's that's a lot of that's a lot of the the work for for these models to to sort of you know have have a have a have have the data in a form that lends itself very very well for the for the training part of it like high quality data for training those models so in some sense once we sort of solve data infrastructure which is the the right query engines and data management which is the right integration then you get an end the right abstraction to sort of integrate and and connect all this data then we get to the next level which is like how do you actually curate it right right and there's certainly very interesting things we can do there as well Mark I think you said it it's a lot of good information I think it's going to be a really interesting sector to follow in the years ahead it's going to be a lot to keep up with uh thank you so much for sharing your expertise today and I hope you come back and talk with us again sometime thank you very much James for having me and what a wonderful conversation thank you

This transcript was generated automatically from the video's captions and may contain errors.