Video: Fireside Chat: Rethinking OLAP – How to Migrate from SSAS to a Modern Semantic Layer | Duration: 3664s | Summary: Fireside Chat: Rethinking OLAP – How to Migrate from SSAS to a Modern Semantic Layer | Chapters: Introduction and Welcome (12.264999s), Introducing the Panel (137.39s), Semantic Layer Evolution (304.43997s), CUBE Cloud Overview (525.81s), Legacy OLAP Challenges (786.81s), Modern OLAP Benefits (1462.78s), Cross-Platform Data Access (2595.925s), Time Intelligence Capabilities (3466.3298s), Conclusion and Farewell (3572.5952s)
Transcript for "Fireside Chat: Rethinking OLAP – How to Migrate from SSAS to a Modern Semantic Layer":
Hey. Good morning, everybody. Looks like we just rolled over and went live. So, looks like we've got folks still, shuffling in here, and, I'll give everybody a couple more minutes, maybe one or two minutes before we get started. So, yeah, thanks for being here with us today. Okay. Well, I've got 11:01 here in lovely central time zone in Oklahoma, so let's go ahead and get started. And, you know, as I said before, you know, thanks for spending your time with us. Good to see you all or, you know, you see me anyway. I don't see you on the webinar, but, yeah, thanks for being here. Yeah. Today's, today's topic is, rethinking OLAP, how to migrate from SSAS to a modern semantic layer. You're gonna hear us talking about that acronym SSAS, because, SQL Server, Analysis Services is is quite a mouthful. So right up front, if you don't know what SSAS is, you're maybe in the wrong place, but, we're here to tell you more about it if you if you don't know what it is. And, it's not just me here today. I'd like to introduce our panel of speakers, that we're gonna have on stage today. So, my name is Brian Bickell, VP of strategy and alliances here at Cube. Been here about two and a half years. Had a lot of run ins with SQL Server Analysis Services in my my prior life, doing data engineering. And, yeah, Artyom, please go ahead and introduce yourself. Good good to be here. My name is Artyom. I'm cofounder and CEO at Cube. Coming from software and data engineering background, I started Cube as an open source project in 2019. And in 2020, my cofounder and I, we build a company around it, build a commercial cloud product and keep building the open source. Excited to be here. Excited to talk more about our vision, how is multidimensional analysis is changing over the course of few decades and how we think about keep place in that change. Great. Mister Michael Treadwell, are you out there? Yeah. Hey, everybody. What's going on? My name is Michael Treadwell. I'm happy to be here with you guys. If you like Artiom, my background is in data engineering. It's a few different organizations in that role, prior to coming here, where I am now the head of sales engineering. So, you know, I oftentimes and and in the first line of of of technical speakers to work with, our our customers, hopefully, like yourselves, to talk about their technical challenges, design architecture, and, you know, help them understand how semantic layers is a, a relevant part of the, the modern data stack. Great. Well, you know, glad to have the, the two of you here with me today. I can't think of a better panel to, to discuss, you know, what we're doing in, modernizing OLAP for our customers. Let's go ahead and, get over to the agenda. So this is, you know, what we're gonna be what we're gonna be talking about today. I'm gonna start us off by by going through, a little bit of history of a little app, the universal semantic layer Cube, you know, what we're doing, and then we're gonna get into the panel discussion. So, I'm just gonna go ahead and get right into it into into my portion of this, and then you will see Artyom and Michael, a little later on. Say say maybe ten minutes or so. We'll see how fast I get through this. Yeah. So, you know, if you're not familiar with Cube or if you if you've if you've seen one of these webinars before, you may see us sort of start off by talking about a little bit of history and then where the idea of semantic layers, come from. Generally, we can trace this idea back to what I kinda think of as, like, the the the monolithic data stack era in, like, the late nineties, like, early two thousands where, you know, you probably got all of your data technology from a single vendor. You know, maybe that's a Microsoft, maybe that's an Oracle, maybe it's a Teradata, something like that. You know, know, for the purposes of today, I'm talking about SSIS and talking about OLAP. Let's maybe zoom in on the Microsoft use case. This is actually the era when I got my start in in data engineering and, you know, saw a lot of, SQL Server 2,000, SQL Server 2,005, softwares. And and, typically, when I saw those with customers, they were tied to ERPs that that customer would buy from Microsoft as well, you know, something like Great Plains. So you'd get a relational database, SQL Server. You'd get, your OLAP cubes for doing multidimensional analysis and doing, you know, roll ups and caching and performance. You know, that was SSAS. You get your integration from them as well and your ERP. So this is all like one one place, to get all your data. But problem with that was, a little bit difficult to use, difficult to bring different people into the fold and get them access to data. And what came after that was what we really call, the era of the modern data stack, which if you're a little bit younger and you're familiar with data, that's that's probably all you've ever heard of. But, that was when we sort of decoupled the data stack into this idea of more of a self-service, ecosystem with a with a bunch of different best of breed players. This is the era where folks like Tableau really grew up, where Looker, sort of made a big name for themselves with LookML and then also with the visualization capabilities of the product. And this is where we saw Microsoft, you know, start to produce Power BI, as their extension of Power Query and extension of Excel. And, you know, what ended up is in this in this in this time period, while we while we did introduce more people into to being able to use data and analytics, we also sort of created a a headache for ourselves in that we created a a lot of mess in terms of, like, where business logic and where, semantics are are located. You know, speaking of the tools specifically, you know, if you have Tableau in your environment, you've got Tableau data sources running around all over the place. A lot of calculated fields and business logic getting cap encapsulated there. Looker, of course, you know, as I mentioned with with LookML, did a lot of this for customers. Power BI has its own concept of a semantic data model. And then if we sort of tie this back to today's topic, you know, in this time period, you also had, SSAS with its concept of, you know, business logic being encapsulated in the analysis services cubes in addition to SQL Server. And that logic probably powered, hierarchies and drill throughs and pivot table capabilities for, a tool like Excel. It's a very, very common use case. The while that's great, and while we gave people a lot of choice, we did create an opportunity for inconsistencies and inaccuracies, and, we we presented an opportunity for people to have to repeat themselves, many times in terms of the, the calculations and business logic. So, what we think about at Cube and really sort of like the the goal of the company and, you know, our idea here is is trying to shift all of these places where semantics and business logic could live, left in the architecture into to one place to establish consistent models and governance, you know, allow you to reuse that model data across many use cases, while at the same time simplifying your access and improving query performance. And, you know, that brings us to kinda what we call the the third wave of of this whole idea of, like, data and universal semantic layers where, Cube Cloud as a universal semantic layer is gonna sit in between all of these different data sources that we're talking about, as well as the, different downstream use cases. And and that's BI tools, that's embedded analytics, that's AI now. And, you know, the idea is to to model data once and deliver it anywhere. If you're not familiar with us at Cube and what the product is, we go to market with a SaaS product called Cube Cloud. And, you know, the big ideas there again are just helping customers unify business logic, make access control easier, provide some speed, and savings through our caching and optimization, and also integrate a bunch of different API endpoints. So, to dig a little bit deeper into the product, you know, we connect to something like 30 different cloud databases, relational databases, query engines. Really, most things that speak SQL, we have a connector for. That's gonna be important to this conversation today because as we see, a lot of organizations that had analysis services or other OLAP cubes, and still have them have left them in place because, they need to, power a downstream tool like an Excel or a Power BI that that that utilizes these connections. But, you know, really, those tools are are best set up to retrieve data from a SQL Server or, in the case of Oracle s base from Oracle itself. They're not really good at pulling data out of cloud data warehouses, so that creates a point of friction. Cube Cloud, of course, you know, we support all major cloud databases. Snowflake, Databricks, BigQuery are probably the top three that we support every day. And in the product, we do, data modeling, that is in a code first but with a visual editor experience. That's all very easily version controlled. We support YAML with, Jinja templating and Python for dynamic programming. We support JavaScript as a language as well. We have a, strong access control capability that handles things like row level security, column level masking, member level filtering, with strong security contexts. So we can kinda handle all of your, security and multi tenancy needs. We have a caching technology that we call Cube store that's built on Apache data fusion and Rust. It's extremely fast. It's very good at pre aggregations and roll ups. If we think about the context of SSAS and and what you're getting in an OLAP cube, pre aggregations and roll ups and fast aggregates were a very important part of those technologies, in the late nineties, early two thousands. And, you know, as data volumes have grown, caching has still become, extremely important, and the ability to do massive, large scale pre aggregations and roll ups and and do them in an aggregate away. Aggregate aware way is very important. It's something that the cube does. And finally, when we think about cube, we have to talk about our APIs. You know, when data leaves Cube, it's gonna leave via a series of different APIs we provide, including REST and GraphQL. That really supports the majority of our embedded use cases where customers are building, customer facing dashboards or embedded analytics on top of us. We also provide a SQL API, which powers, most of our BI tool connectivity. And then, two more additional APIs that are are really, relevant today are our MDX and our DAX APIs, which, again, these are two languages that, SQL Server Analysis Services gives, support for for most customers. And they're the two languages that that primarily power the Excel and the, Power BI workloads. So that's, that's Cube and the history of OLAP in about ten minutes. And, you know, if you're if you're curious to learn more about this, I'd encourage you to to reach out, through contact us or to any of us on the panel. We're happy to dive in and and give you more of the product. But, I wanna get us into the panel conversation because I think that's what most of you are here for. So, let's bring Artyom and Michael back up to the stage. And yeah. Cool. Happy to be here with you guys. You know, I think when we were thinking about putting this panel together, we thought we would bring together a couple folks that, you know, have experience with SSAS and and and all of us have data engineering backgrounds and, you know, we've all bumped into multidimensional OLAP at at some point in our in our past. But, yeah, I'm happy to have the three of you here to kinda help get our audience through this. For our first topic, I wanna start with, what we mean by legacy OLAP, where it stands today, and and what we think is kinda next for this, and and and, really, why is it still relevant? Artyom, I wanna start with the first question, to you. Let's actually just start by defining the problem we're talking about today. What challenges are organizations facing with legacy OLAP platforms like SSAS? Why why bother looking for a replacement? Sure. Happy to start here, Brian. Thank you. And, you already gave a good historical overview where we where we started. And, as an industry, you know, like, how we came up with OLAP servers and, you know, like, where are we going now? I think what what has happened in is that OLAP had its its moment in about twenty years ago, you know, like, when we had all these limitations of OLTP databases, and we needed we needed a better way to run analytical workloads. And we needed a place to build multidimensional models. Right? And we also didn't have, you know, technologies that we have today to to process data in in really fast way. So we we, kind of, you know, like, industry ended up with this idea of a memory storage, like a cell based memory storage that is powering all up technology. So that sort of, like, was this all up server's generation of golden age. We got SSS. We got Hyperion that turned into Sbase, Cognos, all of these technologies. And and they worked, and they they provided they provided value. They became a part of the critical workloads, critical operational workloads in many enterprises for, like, many years. But with Hadoop coming in and then kind of paving the way for for cloud data warehouses, this piece of technology is oh, and when I say this, I mean, a lot piece of technology. It's kind of being in a weird spot right now where it's still being used by many enterprises. It still provides value, but it's definitely on the wrong side of the history because it's not a cloud first. It's not it cannot work on top of cloud data warehouses. In many organizations, they find themselves in a position where they wanna bring all the new new workloads, everything new and exciting into the into the Databricks, into Snowflake, BigQuery of the world. But at the same time, they technically need to support all these legacy workloads inside all app servers. Many of them are being, on premise, right, and not not even a cloud friendly. And we're not talking even about, like, being able to run the Kubernetes server on top of Databricks in a live mode, right, on top of something like in a live mode. So I think that's a problem. That's it's clearly the the, you know, like, the puck is going to cloud data warehouse world. So everyone wanna skate to where puck is going. But this, like, on premise all up servers, they they hold him back. And the enterprises and enterprise data leaders, they wanna solve this problem and to fully modernize and transform their data stack. Yeah. I just wanna continue on with that that idea. I mean, I think, you know, we see that some people that we talk to kinda have this misconception that SSAS still works for a lot of enterprises. I think I think we believe that's sort of a a misleading idea. But what do you what do you think of or how do you think about the hidden costs of sticking with SSAS, whether that's in terms of scalability or in terms of missed opportunity or in terms of pricing or just, you know, what what is the downside to trying to keep the lights on on this on this legacy technology? I'm sure people are gonna think about that because we're, you know, it's impossible to escape that we're in a in a period of economic uncertainty, and a lot of companies are are thinking, like, fewer vendors right now. So, you know, what is the what is the risk here? Yeah. I think the risk here and a and a problem in general as we use any legacy technology. Right? And if it's talking specifically about SSS, you know, like, even Microsoft understand that, and Microsoft wants everyone to migrate over to Fabric and, you know, like, to newer set of technologies, which, you know, it has its own concerns and issues. Right? But speaking specifically about legacy old on premise SSS, the problem is that it the the cost is huge to to still run it and maintain. In many cases, you need to run it on premises in your own, you know, like, a data center. You cannot do it in cloud. And that's sort of, like, an obvious problem. Everyone understands it. Maybe less obvious is that we are coming into newer generation of data professionals and newer generation of data engineers. Right? And the people who worked with this SSS cubes, that generation is going to retire completely over, you know, like, next ten years. And the newer generation, they just don't wanna work with SSS cubes. So it's it's going to be really hard to hire a data professional to to maintain these cubes moving forward. People would wanna move to, you know, like, the new the new technology they used. The more code first technologies, the more modern data stack friendly technology that integrates, you know, across data stack. Snowflake, Databricks, you know, like, some some transformation tools, semantic modeling. If you put a rack out for if you put a rack out for hiring with these technologies, people would be excited. If you put a rack out with SSAS, you know, like, people people just not gonna be excited for that role, and it's gonna be really hard to hire. So that's, like, one of the example of this hidden cost, the hidden risks that enterprises should be aware of. Yeah. That all makes sense. Michael, I wanna bring you into the conversation here. I know that, you know, you mentioned, you know, for everybody who who's just joining, you know, Michael runs, runs sales engineering here at Cube. So he's really first contact with a lot of our customers and and their problems, and he's he's in the trenches every day fighting the good fight. Michael, I'm just curious to get your take on, like, why do so many customers, continue maintaining SSAS? Like, even after they've migrated to cloud data platforms like Snowflake, Databricks, or Redshift. I mean, we've we've been hearing about cloud migrations for a decade or more now. I mean, really more like a decade and a half. And, you know, obviously, Snowflake and Databricks are great partners of us, and they've they've grown by leaps and bounds. So how how is there still SSAS out there? That's my question for you. Yeah. I mean, RDM covered a lot of the technical pieces, right, of of cloud's interoperability. But, you know, it's a good question because if they are interoperable, why are we still dragging these things along in this in this parallel world? And, you know, from what I see, the reality of the situation is that users, analysts at organizations have specific working patterns that bring a lot of value to the organization. Right? And those working patterns aren't necessarily supported in, like, a first class way using only a SQL interface, right, which is what Databricks and and Snowflake and the cloud data warehouses of the world, want to want to operate. And and, you know, Brian, I know you've seen this as well. Right? We pretended that spreadsheets would just go away for a long time, but they're not. And they're not for a reason, and it's because the the work that people do on them is it's still bringing value to the organizations. Right? So and, yeah, you can connect Excel to a database, but you don't get the responsiveness. You don't get the performance. And in many cases, you don't get the functionality that you need. Right? Excel doesn't write SQL well. It's not what the tool was meant to do. Right? It's it's a meant meant to, supply pivot, cross tab spreadsheet analysis, which is still the core of how a lot of people think about data. It's the core of how a lot of people still work with data, and it's the core of how a lot of organizations still provide value, from their from their analytics organization. So ultimately, you know, what we've realized here at Cube is the answer isn't to tell those users that their working patterns are bad. Right? It's not to tell them to go away. In fact, one of the great things is that, you know, I've received feedback from customers telling me that it's refreshing, to see someone treat MDX and DAX as first class instead of a problem that we just wish didn't exist. Right? Because if you wish it didn't exist, it's still going to exist. It's just gonna exist sort of in parallel, to all of these these other systems that you're running. So I I think that's the biggest thing is, you know, we we have to we have to consider the future, that the future still evolves, that it still involves spreadsheets. Right? And we still have to support that even when when everyone's moving to the cloud. So we have to have a modern way of doing that. Yeah. I think, you know, I mentioned in the beginning in the intro that I've I've been here, you know, quite a while with the business and and seen the evolution of Cube as a product. And I think one of our our north stars has always been giving customers optionality in terms of, like, what they wanna put on top of Cube and how they wanna build their stacks. And I think if you've been in BI for any amount of time at all, you realize that BI is really accretive. We don't ever get rid of tools. We just keep adding tools, and it's like cities built on top of other cities. It's kinda like human history in a weird way. And I think that what a lot of, like, new vendors in the space sort of misunderstand about Excel is they say, I'm gonna build Excel for the cloud. And they they don't realize that, you know, you're not gonna get customers to throw out thirty years of workflows. That's at least my take. I mean, I I wish them well, but we we try to build for a little bit more, pragmatic customer. I wanna move on, and we're gonna get to kind of the second broad topic here, in the, in the panel today. We're gonna talk about the the case for modern OLAP and the semantic layer. I know in a lot of our marketing, you know, we use this phrase, modern OLAP, trying to tie, the history of SQL Server Analysis Services and Essbase and, you know, bits of SAP HANA, Team One, other OLAP systems, Yeah. To what we're doing here at Cube, with the universal semantic layer. Artyom, I wanna address this question to you because, you know, obviously, you've you've been behind or or in in a lot of the conversations on on bringing this to to bear as a product. How does the universal semantic layer solve the challenge that SSA as users face today? I mean, how do we how do we kinda get up here and say it's a it's a it's a replacement that makes sense? Sure. First of all, how do we define semantic layer and what is that? Right? We we we kind of call the company cube for a reason because, you know, like, obviously, it's it's multidimensional cubing technology. Right? It's different from OLAP. It's different to SAS. It's different from that generation of cubing technologies. But at the end of the day, it's it's multidimensional analysis. Right? So the semantic layer is built to allow data engineers, data stewards to create multidimensional models. If you if you go on our website, we have this kind of four layers of functionality, and the first layer is foundation of everything is multidimensional data model. Right? That's a place where you can build your cubes, define relationship, build dimensions, measures. So from that perspective, we offer the feature parity to tools like SSS. In a way, when when you can build an SSS, you should be able to build in pure bright all the all this complicated, multidimensional analysis, metrics, you you know, like, hierarchies, dimensions, all of that. But then we do it differently from a sort of like a product design in a way that we run on top of live data in in cloud data warehouse. I think that's one of the major differences. But we also wanna make sure that we are interoperable as much as possible. So instead of kind of designing the system multidimensional system that's supposed to work within one closed ecosystem, and, you know, like, that's the case for many of the existing co op technologies because they're owned by bigger vendors, like SSS, part of Microsoft, and it works well within Microsoft. Right? You know, like, Oracle has one. IBM has one. So they all work and were very well integrated within their respective ecosystems. But when we started the company, our goal was, okay, can we make it open and interoperable so it can integrate with existing data consumption tools? But, also, we're just just an ecosystem in general and, you know, like, systems that you can build. If you wanna build your own front end with React, you can do it. If you wanna build your Python app as a front end, you can do that. But that's sort of, like, the design principle we add from the first, you know, like, from the beginning is that how do we make it headless and API first. Right? So think about it then and why it can replace all up and why it's better is that it has the same power of multidimensional data modeling, and it also gives you a set of APIs and endpoints that you can connect Excel, Power BI, your React application, your PPI, and all these tools. So you can kind of push data into into all of these places. And on top of that, and we probably will talk more about it later, but we do create data caching, extraction, all of this stuff completely different, not an s SSS. Right? We're not, like, in memory. We're not requiring users to extract data out of Snowflake all the time and wait for this extracts to rebuild. So that it works Cube works very differently here, and I think that's sort of, like, our unique difference as well. Yeah. I'm gonna stick with you just a little bit longer, Artyom. I know I know you kinda mentioned, you know, how we think about the replacement. But can you maybe elaborate on some of the more of, like, the the cloud native, benefits over traditional SSAS OLAP? And I I'm really thinking more about, like, some of the, the scale that we've seen in our customers in terms of, like, Cube ability to scale horizontally. I know that we mentioned before that, SSAS is is, you know, stuck on a Windows environment. It's usually stuck, in its amount of horizontal scaling. It's typically scaled vertically, because it's like one system. Yeah. I mean, can you maybe elaborate a little bit on, like, the horizontal scaling aspect of caching? Yeah. Yeah. Sure. Well, maybe maybe we should talk about the caching first and, you know, like, how it's done and then, like, expand a little bit of, like, how does it scale. So in a way that, you know, like, all up additionally, you know, worked and SSS, specifically, it's all in memory. Right? So to to get a bigger cache, you need to vertically gather just a bigger, you know, like, machine with more memory. And then you you kind of put in your extracts into that. The way that Cube works, the Cube kind of cache works and a Cube in general is all cloud. Right? So we run Cube on AWS, Azure, or GCP. So we can run it in our accounts. We can run it in customer accounts, but it's all a cloud based. And, the way that Cube Cloud works, it's a distributed compute engine, Similar you know, like, it's a big data, essentially, like Hadoop AD is MapReduce. So you get a bunch of worker processing nodes, and you have routing nodes. And the this way, you can horizontally scale the processing and the to be stored the data not on, in memory, but be stored on SSD disks as as per k files. So this gives you, you know, like, horizontals horizontal infinite scale, essentially. That's one kind of the benefit of that. But the other benefit is that you can incrementally update your cache, meaning that think about it. You have, like, this historical data for, five years. Right? You don't need to rebuild this five years extract every time you make a change. You can only change the last months, right, like, once you add a date or something like that. So you can incrementally update update your data. So this benefits from a scalability perspective. They really make a difference compared to older, like, in memory of products. Yeah. I think, you know, I know from talking to some customers, we we've seen some scenarios where, customers that still have legacy SSAS are are running into, like more than twenty four hour rebuilds, which really kind of blows up the pace of business if you're trying to do a daily export and you can't you can't rebuild the cube in in sub twenty four hours. And that's kind of stuck because of the, the inability to scale vertically beyond systems that have, you know, whatever memory they need to have. Michael, I wanna I wanna kinda bring you back here and and maybe come back to everybody's, you know, favorite topic, which I I think is is increasingly everybody's favorite topic at Cube is is how how Cube interacts with, with Excel. I know when people are out in the audience and they're kinda listening to us talk about us being a SSAS replacement, they're they're really gonna wanna double click on how Cube Cloud, maintain seamless Excel connectivity. So can you kinda just explain the different the different approaches we have? Because there's a there's a few different routes. Right. Yeah. I I wanna start off with a bit of disambiguation because I, you know, I I think that, oftentimes, tools say, well, we connect to Excel or we connect to Power BI. And what they mean is is is one of two things. One, they mean there's some third party implementation, you know, either some third party tool or third party implementation that you have to acquire and get approved in order to bring in to, to make that happen. Or they're just in some sort of direct query mode where you're letting Excel or Power BI just directly, create a query, which is not ideal because, again, like I said previously, those those tools really aren't designed, to create queries, SQL queries. Right? They're not SQL writing tools. So, yeah, I mean, we have two primary ways, from Excel that we that we connect, to the to the semantic model. The first is, you know, ex an Excel add in. So if you're using Excel online or if you're, you know, intro if you're accustomed to the workflow of, like, SAP HANA or some non SSA solutions, you know, we can we can provide an add in for you, which gives you the ability to natively drag and drop and sort of, like, cubish interface to produce those queries. But what's more impressive, is our is our MDX capabilities. So, you know, those of you who have, you know, gone one level down on Excel, one level down on on SAS SSAS, you'll you'll understand that, you know, Excel speaks, to SSAS, in MDX. Right? Multidimensional expressions. This lets you query multidimensional objects like you would get out of, out of something like SSAS. Right? And the interface for that, that SSAS uses is XMLA for analysis. Oh, sorry. XML for analysis. Right? XMLA. It's protocol, to, you know, facilitate that communication. And so rather than trying to bypass that, we or, you know, trick the system, we've just gone directly into that so that we can provide native Excel functionality. Right? So when you work with Cube, and you build your models, we provide you an XMLA endpoint, so that when you're working in Excel, you connect to Cube the same way you connect to SSAS. You you you even choose the analysis services option. You you enter an XML endpoint that we provide, and Cube sorry. Excel provides MDX to Cube, and we interpret that directly as a SQL, according to the according to the model that you have built. Right? And and and this MDX piece is very important, because I'm sure, if anybody is on here who supports it as a SAS deployment, it's very clear, you know, where where Microsoft is taking this. Azure analysis services, the fabric model, the future for those are tabular models, right, not MDX. This it's this this much is laid out very clearly inside of inside of Microsoft def Microsoft's documentation. Right? You can still run MDX cubes, in Azure VMs, if if you if you really need to get that workarounds. Right? But what we're saying here at Cube is that multidimensional analysis is still relevant, and that there's a modern way of doing that. We are enabling that with the first class integration, without any any weird, third party workarounds, to accept connect Excel and Power BI to cube, as if it was, an SSAS cube and provide the performance that you would expect via our pre aggregation cube store technologies, that that RDM just covered. Cool. I really appreciate that, Michael. I think that's a great overview of, of Excel, and I I know that's always on everybody's mind. So, in the interest of time, I wanna move us to, what is gonna be kind of the the last big big topic in the panel, and we are gonna have some time for a demo, and then I'm gonna try to get time for q and a here. So, I know we've sort of been at this pretty, abstract level. Right? But maybe let's let's get down a degree into what, you know, migration strategy and, like, real world adoption looks like. I can tell, you know, looking at the the chat and the QA, there are a lot of people, here attending that have, like, real, you know, questions about how this works. So, you know, let's get into get into, you know, how what it's gonna take for a customer to get there. Michael, I wanna address this first question to you. You know, these customers of ours that we're that we're talking to these organizations, you know, they've had SSAS for, you know, a decade or more. I mean, you know, probably two decades in a lot of cases. It's really a massive undertaking to to start to think about a migration. How do you think about that from the sales engineering perspective? And, how how should organizations approach that transition in a way that minimizes their risk and and gives them the the best upshot? Yeah. For sure. You know, I'm not gonna I'm not gonna sort of beat around the bush or, you know, play play games here. Right? This this can be a this will be if you have a a large SSAS deployment. Right, it's gonna be a large undertaking. Right? There's a lot of history there. There's a lot of business logic there. And it could be a a lot of work. Right? So in in some ways, you know, it can be a little bit like a war and, you know, you you you gotta have a you gotta have a really good plan, is is the thing. Right? And so we're here to help you with that. You know, a lot of that stuff will be uncovered as part of our POC process. So even by the time you finish the POC and understand the Cube is the is the tool for you, you will have some of that, as well as our our support team and our support partners. Right? But, you know, there at a high level, there are three things that that we we suggest. Some people I've heard, I I didn't get this term. I I got it from one of my customers who was doing something like this. They called it a listening tour. But what you wanna do is you you have to document your users' workflows. Right? Because that was the first thing I said. The first question that Brian asked me, my first answer was, users have specific working patterns that bring a lot of value to organizations. Right? And none of this is gonna work unless you have their buy in. And in order to get their buy in, you and understand how they bring value to the organization, you have to understand how they work in these tools. How are they using the cubes? What tools are they using inside of Excel? Which specific pieces of functionality that they're using? Right? We have to understand this at a high level. The other piece of the plan, is to understand the the reach that each of their cubes have and the analysis that they support. Right? So before we go changing something, we have to understand what we're changing. We have to understand the reports that that affects and thus the decisions that those cubes affect, and and, you know, the the reach ultimately, the number of of reports that are built on top of each one of those. And then the last thing is is a timeline. Right? So once we understand what we're what we're doing and how we're doing it, we have to have a timeline. Right? And, again, a lot of this comes out a lot of this is uncovered, for free, as part of the POC process and is documented, as you're working through. But, yeah, I mean, just the the planning. It's it it really it, you know, there's a lot of modern solutions, but there's no button you can press that just, makes this automatic. So having a plan, having the training, having the documentation, having the understanding of the impact, and then moving port before even before it is is critical. Yeah. I mean, to add to that, you know, I think having done a lot of migrations in the past from legacy BI to, you know, what's now still legacy BI, it's It's how old I am. You know, a lot of times it's inventory, and it it is it is just accepting that you don't have to move everything. You know, we talk about these systems being accretive and being in place for ten, twenty years. There are gonna be some stuff that's no longer business relevant. So so do your inventory and and get really tight on that and then just figure out what you have to move to be successful. I think that's the most important part. And also, I run alliances here, so I'd be remiss if I didn't mention there are a whole lot of consulting partners at Cube that can help you with that. So, we have a lot of great services partners out there that are very, very educated on this topic. With that, I kinda I wanna move us out of the panel portion. We've got a lot of great questions that have come in the chat, and, looks like we got about twenty minutes left. So I wanna get us over to, you know, talking about what this looks like in the in the real world. We did have one more slide here. I wanna present this, briefly, and and this is just, you know, what we think about at Cube, as the road map for the different, BI and spreadsheet relevant APIs we have. I know it's been maybe a little bit hard to follow the conversation if you if you don't kinda have this, this map here. But as I mentioned at the beginning, you know, for traditional, BI tools, we have a SQL API. For Power BI, we have that DAX API. MDX, as Michael said, supports Excel. We also have a Cube Cloud add in for Excel for, for platforms that can't speak MDX because of drivers or or something else. That's macOS. That's, that's Excel online. And then, you know, germane to multi dimensional analysis, we also have a Cube Cloud for sheets connector. So So you can do this kind of pivot table analysis in Google Sheets as well if you're a shop that's, you know, doing that instead of Excel. I'd point out these are all the same, the same data models, all the same caching, all the same governance, and that's really the the the big power of Cube. With that, I want to get us into the demo. So, Michael gonna take us through how this works. We have done a lot of talking about how cubes universal semantic layer can be a modern replacement for your legacy SSAS OLAP Cube implementation. However, we don't wanna just be all talk. We want to show you something. We wanna show you what it's like to use Cube and what it's like, to connect to Cube from Excel, from Power BI, the Microsoft tools, right, like you would be using SSAS, but also how you can share that model that you have built in one central place, something like Tableau. So let's take a look at the visual representation of the model that we are going to be working with today. So this is our TPC CH demo data. This is a popular demo dataset that many use people use for benchmarking, and we are using the TPCH 100. For those, if you know who know what that means, it means it's 100 times the scale. So we do have large data that we are working with, not just a tiny dataset. And this is a line item fact table with some order and part in region dimensions that are attached to it. And all of this is built out inside of my data model. So here is my line item cube. This represents my fact, and we're joining the part supplier as well as the orders cube. We are defining these joins here so that if they're needed in order to produce the SQL, they can be produced by Cube. So we're defining those relationships here. We are defining some dimensions, which are just pull throughs of columns from the underlying table, but we also are adding a new business logic here inside of our data model. So, for example, with ship date, you can see that we are defining a fiscal year and a fiscal quarter, these custom time granularities, so that our users don't have to remember which month or how to use the date add functionality when they want to refer to our fiscal years additionally. We're defining some case statements that return labels instead of flags so that your users can see in plain words what the flags mean instead of trying to guess what those interpretations are. So as you see, we're embedding our business logic, our joins so that users don't have to understand the granularity of the tables. Our our dimension business logic, the the math that we use to define things like revenue, we're embedding that inside of the business, inside of the data model itself so that users can simply query the logical form like you would with something like SSAS, all of the measures, the metrics, and the dimensions that conform to it rather than than having to write SQL in the joins themselves. So in addition to just, you know, defining revenue, we're defining these year over year calculations as well as some year over year comparisons to get a percentage value of what our revenue this year is compared to our prior year. And then because these year over year calculations, you know, these these can be, you know, pretty intensive. Just like with an SSAS cube, we are rolling these up into a roll up that Cube calls pre aggregations, and this will serve as the the the pre aggregation, the caching layer inside of Cube, which will take the place of your SSAS cubes. And here you can see we're defining this as part of our data model. We're rolling up to this late level of dimensionality, and we are rolling up, these measures to that level of dimensionality. And, additionally, we're applying some other logic here, some partitioning so that we can rebuild these incrementally, which you can do with pre aggregations, as well as indexes so that certain query patterns, are faster when querying this preaggregation as well. So any of these pre aggregations that you can that you build, you can see here inside of our pre aggregations page. And this one in particular that we are working with is our line item inside of our line item Cube, and you can see the yearly partitions. If you wanna come in here, you can build each one of these incrementally, or you can schedule them to rebuild, or you can trigger to them to rebuild as you see fit. So once all of this is built, we roll it up into what we call a view, which is where we define the dimensions and measures from each Cube that we want to be inside of our view, and then we expose that view, in this case, the revenue margin view to the user so that all they have to do is logically pick and choose what they want. They just pick revenue. They pick, customer region, customer market segment, whatever dimensions and met combination of dimensions and measures, and then Cube based on the data model that you have built, writes that query and either sends it to our pre aggregation or it sends it, to your data source. Right? So with Cube, you can get the best of both worlds. You can, serve queries automatically off of preaggregations off of cubes. But as soon as you drill down past the level of the Cube, it will federate to the data source so that you're not locked in to a specific level of dimensionality like you would be with SSAS. Okay. So, let's before we jump into Excel and Power BI and Tableau, let's take a look at what query this looks like inside of our playgrounds interface here. So we're gonna just put together a simple query like supplier region and nation, And then we will add in the revenue metric here, and you can see this returns really quickly. We're gonna see up here momentarily how long that took. It took ninety one milliseconds off of our pre aggregation. You'll see that we're only seeing data here for America. That's because the security context I'm passing here is James at cube dot dev. And James, because of the row level security that we have set up inside of our model, only has access to the rows for supplier region America. Now when we access this data with Excel, with Power BI, with Tableau, you'll see I'm logged in as myself, Michael, at cube dot dev. So we will have access to a different set of rows, specifically Europe and The Middle East. So let's go ahead and access that and query it directly. So we're gonna open up a brand new workbook here. And then Excel, we're going to access our data just the same way we do with analysis services. That is because Cube provides us with an XMLA endpoint that we can utilize here. Now you can set up Windows authentication if you'd like. This demo environment that I'm using, I'm using my username and password directly, but I'm still being authenticated. Check my password. I can access the revenue margin cube, the view that I have created, and then begin building a pivot table with that. So you should see the same dimensions and metrics that I have made available inside of our cube, and let's just really quickly build out a visualization here. So we'll choose supplier region and nation just as we did inside of the playground. But one quick thing to notice here is what? I'm getting data for Europe in The Middle East because I have row level security applied. So when we build this row level security inside of Cube, it applies not just to our extracted data, but also any queries that federate back to our data source as well. So it's one centralized place to manage access. Let's go a little bit deeper with this. Let's build out a, crosstab here. We'll put order date on our columns, and then maybe we can filter by brand as well just to show you that we can get that snappy behavior with our pre aggregations that you expect from SSAS. And this is built on quite a bit of data. There are, several million rows of data here, that we are pre aggregating, and this can get much, much larger Cube scales horizontally. So if you need more capacity, if you need more Cube store capacity, you can just add that if you need to build more cubes and add more data to those cubes every day. K. So we don't wanna just build this in Excel. We, of course, want to access this with Power BI as well. So we will get data from a source, and we will get that from SQL Server analysis services. And, again, we provide an XMLA endpoint or rather a DAX endpoint that we can connect to directly. We will once again choose our revenue margin view that we have created, and that data will become available to us over here. And let's just, in the interest of time, build out a simple bar chart visualization here. We can add something different. Let's do, maybe, customer region and then revenue, and let's do that by customer market segments. Alright. And then we can see these getting built out here. Right? And, of course, we could add filtering and whatnot. So we see the same set of metrics inside of Power BI as we do with Excel, and it's not just with our Microsoft tools. Right? We can access this with Tableau as well. So using Cube semantic layer sync, we automatically build data sources for you every time you change your model. So we can just come here and build a workbook using this data source. And just as with Excel and just as with Power BI, you will see the dimensions and measures that are available to you. And, again, you'll see my real level security working here. And so all of these queries that are being generated are coming back to Qube because we connect to Qube instead of our data source, so we are, of course, able to see all of these queries here. You can see most of them are being served very quickly off of our pre aggregation. That's what the brown lightning bolt means. And you can see that Cube is directly receiving and interpreting DAX, from Power BI. We do the same thing with our MDX API as well. We don't have a middleman that we work with with DAX and MDX. We receive and interpret those directly without any third party tools that are required. Then, of course, you see the queries that Tableau is sending as well. So you can always see the query that Cube received as well as the SQL that was generated. In this case, it will be the SQL that was generated against our pre aggregations. So that's how building a single model inside of Cube and then exposing it via our DAX, MDX, SQL, REST, GraphQL, or AI components works. Great. Thanks for the great demo, Michael. I think, you know, the story of how the Cube supports, you know, not just MDX, for Excel and DAX for Power BI, but but also, you know, traditional BI tools like Tableau, the same data model, is is is hugely, important. So, yeah, glad to, glad to be able to show the audience that. We're gonna go ahead and move into our, q and a. Looks like we've got about five minutes left in the session here. So I'm gonna dig through the chat, and I'm gonna find some questions, and we're gonna get them answered for you. And if I don't get to you, you know, you can definitely reach us at the, Cube dot dev and then the contact us form. Or you can find you can find Michael or I on LinkedIn or however you wanna find us. We're easy to find. Let's go ahead and let's go ahead and take a look here. Let's see. Okay. I'm gonna start with this one, Michael. We've got a got a viewer out there that asks, for clients with multidimensional cubes with fact tables over a hundred million records serving Excel reports today, what is the Cube skew that is necessary? And then he also asked about performance. So and do you wanna talk about, like, what what version of Cube this is available in, you know, for access to the MDX API? Yeah. As far as SKU goes, you know, you will be utilizing our our at least our enterprise tier. Right? If not our enterprise premier tier depending on your deployment model. So so that's what where you'll be positioned. As as far as, you know, a hundred million rows for the cube, you know, row count is obviously important. Number of dimensions, so the width is is is important, as well as, obviously, the the cardinality, overall that it combines. But I will say, you know, we we have a lot of organizations that have, you know, pre aggregations with a hundred million rows. I believe the one that I was showing very quickly, was is is, what we consider a small one. Right? And they had about 20,000,000, in the the demo that I was just doing. You know, and as Artyom said earlier, cube scales horizontally. You know, we we are a modern cloud technology. So adding cube store workers, adding partitions, adding indexes, you know, Cube gives you a lot of, capabilities that SSA is not in order to manage more volume, larger datasets, and update those in a sensible way. Nice. Couple more minutes left. We had a few people here ask the same question, wanting to know about Cube ability to support, time intelligence in DAX. So, Michael or Artiom, do you wanna take that question and and maybe talk about, you know, what we do on on that topic? Yeah. Sure. I I can grab this one. And, again, in the interest of time, I can also point people to some documentations. You know, Cube supports, custom time granularities. You can embed those as part you define those as as part of your model. And, additionally, you know, we we do provide, time intelligence as well, both from our rest API and our SQL API. So, you know, dynamic query rewriting is is the core of what we do. So so managing that as part of the model and also as part of our SQL interpretation is all possible. Yes. Nice. There's a question Go ahead, Artyom. Yeah. Go ahead. For the time intelligence specifically, there is a concept in the cube called multistage where you can build or, like, it's possible to create a metric that would require making a multiple kind of calls to your underlying warehouse and then calculate the the derived metric out of that. So that could be used, for things like, prior months revenue that, you know, like, you can use to build, growth over time. That's, like, a very common example from a time intelligence. So from the documentation perspective, search for multistage metrics, that would be the best place to start looking into time intelligence metrics. There's a few folks in here with general, like, scale, you know, kind of questions. You know, one one user asks about having cubes with, you know, over 200 fact and dimension tables having millions of records. Michael, RDM, any concern about that from a scale perspective? I I think the answer is no. But, I'm I know based on what we've seen already. Yeah. The quick answer is no. Right? As part of the POC, we'll we'll understand your volume requirements. We'll understand the required dimensionality and the refresh rates, and, we will design your your cube environment to to support that. Right? Cool. Looks like we're at the top of the hour. So, if I didn't get to your question in the chat or in the q and a, as I mentioned before, track us down, cube.dev, and you can use the contact us form. I'm just brian.bickel@cube.dev. If you wanna send me an email, I'm easy to find. And, yeah. Happy to, happy to answer any other questions. And, we will post the recording and, we hope it's been interesting enough. You'll send it to all your teammates. So, yeah. Look for that, after the session. Thanks, everybody.