I continue on with my Interview series with Analysis Services Program Manager Kasper de Jonge (Blog|Twitter). As before, some edits were made, with Kasper’s permission, to eliminate byproducts of casual conversation and make things flow better in writing.

Transcript

Mark V:

How would you say my job as an SSAS developer would be different in five years?

Kasper:

Before I joined Microsoft, I was a developer, myself. I developed Analysis Services Cubes and SSRS reports on top of them. And they never seemed to work very well together. One of the things I have seen over the years, since I joined Microsoft, is the Teams started working together better, much better. So, teams like Power View and Analysis Services are coming together in releases, and now Power Query and the Data Steward experience join the mix. But I think that is one of the key aspects going forward.

I have been trying to sell MS BI before joining Microsoft, and it was hard. What do you need if you want to buy MS BI? You need Excel, so you need an Office license key, you need SharePoint, you need Analysis Services, you need Enterprise Edition, or BI Edition now, luckily we have that. So, you need to sell four different products. Now you can just say, we have one product: Power BI.

It’s gradually going. Power Query is still a little bit separate. The M language is there, then there’s the DAX language, and what do you do where? But at least we’re landing. The first thing we said two years ago was that there’s only going to be one model. And that’s the Analysis Services model. In the past, Reporting Services had their own model, right? The SMDLs [Semantic Model Definition Language]. Performance Point had their own models. They all had their own stuff. So we said, “No More. There’s only going to be one model, and that’s going to be Analysis Services.” That’s already a big step. You see people like Power Map come into the picture. The initial versions that were not public were not really connected to our stuff. We sat down together and said, “Let’s be sure we all do the same thing.” So, if you go into Power Pivot, and you say this column is a country, tag it as a country, not only can Power View use it, but Power Map will now use it as well. I think that’s one of the biggest benefits and it was really needed: to make one product, and make them work much better together.

Mark V:

Do you see big changes in the skills of people like myself, not an Information Worker, but someone who sets up the environments in which the Information Workers play?

Kasper:

I don’t really think so. I think the role is going to change a little. And that’s not necessarily to say that you’re going to have to do different things. But in the recent years, as there’s less IT, more cutbacks in IT, you have to do more things in less time. So, enabling the Business User is becoming more and more important. And not just by giving them canned reports, but by giving them better models, which we already did with Multidimensional Models for years. But make it even easier, and that means making good models in either multidimensional or Tabular, and have a good analytical platform on top of that. So, that’s one kind of user who only wants to do template reports or ad hoc visualization on top of models. That kind of stays the same, I think. I do hope that with Tabular models, it’s becoming easier to do shorter iterations, and we can grow the Tabular model over time and make it easier to use and make it easier to do larger things. For example, I have seen people that have six to seven hundred measures in their Tabular model. And that’s pretty hard to maintain. So, we need to come up with stuff to make that easier. I met someone yesterday that had 120 tables and five hundred measures. Well, right now, we don’t have a great experience for you to build and manage that. So we need to think about what that means. It’s more about how the tools change. I’m a PM [Program Manager] who works on the tools side of things. So, that is one aspect of the BI Pro as we know them today.

On the other side of things, with data movement, as Matt Masson was showing earlier today, you can expose data for your users to start using inside Power Query. And you can enable data steward to start creating data. So, you, as IT, are not necessarily building it, but you are starting to enable people. And I remember, back in the day, when I was building cubes myself, I built an application in .Net that allowed business users to add data to the data warehouse. Master Data Services does it now pretty well. So, the two types of Business Users, one being the user that just wants to do reporting, doesn’t want to do any modeling themselves or any calculations. So, that’s one. The other is the actual Power Pivot/Power Query user and we can help them get to the right data easily and make them confident that the data is right. And that’s an important venue. And I think that’s also an important part BI pros have been doing for years. They can shift a little bit into that mindset, and enable that as well.

Mark V:

From a tools perspective, one of the questions I have around enabling the end user to get more and more data, including data directly from the Internet. One of the things you talked about is the experience for the data steward with Master Data Services. Is there discussion around a solution that allows users to get data from the internet, but only so much. Kay Unkroth, at TechEd, did a great session around Managed BI. In that session, a fictitious company tracked the purchasing habits of Pink Panthers. And it wasn’t until a large investment had been made that someone realized, “Oh no. Pink Panthers aren’t real.” So, the experience of getting to more data. But how do we make sure it’s good?

Kasper:

There are definitely discussion about all of that. And you already see it a little bit in the portals. If you saw Matt Masson’s session today, you saw that you can track how many times different data has been used, and by whom. And we have that On-Prem today. And, in my mind, that is one of the most popular things. To allow you to understand what the data means. And I sincerely hope, and I am not sure if this is coming, but things like Data Lineage would make a lot of sense in here as well. I don’t know if you’re familiar with Prodiance? That’s something that the Excel team has. And it’s already released in Excel 2013. And it allows them to do, sort of, Excel spreadsheet lineage focused on the financial markets. I don’t know if you remember, this was a few years ago, and someone made an error in some calculation in an Excel spreadsheet and they lost a few billion dollars. So now all banks, etc, are saying, “OK. We need to manage this.” So they [Excel Team] have a product they bought, I think two years ago, called Prodiance. And it’s now available inside Excel. They only discover Excel workbooks for now and they don’t know anything about data models and everything that goes into that. So, it would be great if we could “hook that up” for example. I’m not saying that we’re doing that. But that’s something that would make sense.

Mark V:

So, with the way that Office and Analysis Services are dovetailing more, like in Power BI, is there sometimes contention between the teams?

Kasper:

No. The Office team loves what we’re doing. We’re adding value to Office. We’re giving them all kinds of new features. And we’re innovating in the BI space. And they love that. They do give us some hints and tips on what they want to see and we try to accommodate that. It’s more like working together. Our directors are working together and they see what is needed and say, “How do we work together on doing this?” We all see we’re working together in the Excel code base. But what do you think about Power BI? It’s one completely shared code base. You have Office 365, SharePoint Online, all the infrastructure. It’s one big surface that lives and breathes together. So, it’s a lot of working together.

Mark V:

That has to be pretty exciting.

Kasper:

Yes. I mean, it’s a big company. The Office team has its own building. It’s a little bit different. Each team has its own rules, and how it works, and it’s different. Office has a longer planning period. We don’t have a long planning period. In the past, we also had different shipping vehicles. Now this is more streamlined.

Mark V:

So, with the evolution of Analysis Services to feature both the Multidimensional and now the Tabular model, I encounter people who say, or have heard others say, “Multidimensional is dying” and “Don’t bother learning MDX because it’s not going to matter anymore” and so on. What kind of message would you have for those people?

Kasper:

My next session, in an hour, is about all the investments that we made in multidimensional that allow you to do Power View over Cubes. And that was not easy improvement. So, we now support DAX queries on top of Multidimensional Cubes. That is some major major work that has happened. We’re saying, now you have all the good stuff with Power View. And whenever Power View does something going forward: you will get it. Automatically. So, it’s definitely not that. Having said that, it’s still a hard decision on when to go for what. Multidimensional is just a much more mature product. It’s been in the market for so long. People have worked with it for all these years. With Multidimensional, we’ve seen all these different usage types.  We’ve seen the Yahoo cubes, the huge ones, the small ones, we’ve seen people do Writeback, and all those kinds of things. So, it’s been around the block. Tabular has not been around the block for long. It just started the journey. So, we’ll see where that ends up. I’ve heard some feedback from people here as well. They did multidimensional cubes and they started Tabular and said, “Well, it’s just great because it makes it so easy and makes it so fast to build something.” But it doesn’t have certain features. That’s for sure. Calculated Members would make my life so much easier. I wouldn’t have to do 400 measures. If I have Calculated Members, I could just have a few Calculated Members, and I’m done. I don’t have to do YTD for this measure, and this measure, and this measure. And when I do custom rollups, you can’t do it in Tabular. There’s just some things in Tabular that you cannot do yet. For example, Hierarchies. Get me the Parent of something. In Multidimensional, is makes sense because you have Attribute Relationships and you have Hierarchy structures. In Tabular, we don’t. We just have Tables. We have Hierarchies there, but hierarchies are more an “ease of use” feature instead of a structural feature, like it is in Multidimensional. So, there’s just a lot of things that haven’t made it. We don’t know if we want to bring that in to Tabular. So, it’s not that, that’s for sure.

Mark V:

Multidimensional is not going away.

Kasper:

No. It’s certainly not going away.

Mark V:

So, with MDX being as complicated as it is, and even though it would take years to get really good at MDX, is it still worthwhile path to go down since there is still so much multidimensional out there?

Kasper:

Yes.

Mark V:

And there are still so many use case for Multidimensional, even with Tabular.

Kasper:

And Excel still talks MDX even to Tabular. There are so many tools out there that talk MDX. But, having said that, I’ve heard a lot of people here that said, “I’ve migrated a lot of Multidimensional Cubes to Tabular Cubes. It makes my life so much easier.” So, I’m not sure I can give an answer. But, I think you can get away with just learning the basics of MDX. Or learning the basics of both. Because, I think, that’s probably what you’re going to need. You probably think about, “What do I need to become an expert in?” I’m not sure what the answer is.

Mark V:

It’s kind of tough. That’s the position I’m in, personally. I’ve done a little MDX. I have a blog series and stuff like that; went really well. And I’m like, “Well, do I dive deeper into that? Do I do something similar for DAX?”

Kasper:

It kind of depends on the situation you’re in, I would think. If you have the opportunity to push Tabular, it fits much more into the Agile world. I mean, it’s so much easier to make some changes. But, if you’re customer demands are not Agile, if they want to stick to the old world methods, then Multidimensional is probably preferred, I would think.

Mark V:

So, having been on the [Analysis Services] team for a few years, are there features of Tabular, of Power Pivot, or anything that you championed and are really proud of? Anything where you’re like “Hey, I stood up for this, it’s in the product, and I’m really pumped?”

Kasper:

It’s so much of the little things. I have business, myself, with everything. Thinks like this particular DAX function; I need to make sure this works correctly. All the small things like Sort by other Column; making sure that came in.

Mark V:

I love that, by the way.

Kasper:

It’s so many of those little things that make the product complete.

Mark V:

I did a POC for a client using Tabular because it’s really a good fit and it was kind of a cool solution. One of the things I found when I was working on it was that, working within SQL Server Data Tools…. It’s not “awesome.” You can do it. You’ve seen some of my Tweets about changing Column names and things of that nature. There was a great tool that Cathy Dumas had written and put on Codeplex.

Kasper:

The DAX Editor one?

Mark V:

Yeah. The DAX Editor. Are there any thoughts to maybe upgrading that? Because, even though it was not fully compatible with [SSAS Tabular] Service Pack 1, and it had “issues,” it was awesome enough, that I used it anyway.

Kasper:

That was a personal prototype, together with someone else. I cannot speak for that person.

Mark V:

OK. But something like that. Writing DAX in THAT environment, even with it not working perfectly, was awesome.

Kasper:

I know. I get that. They found a quicker way to do it. Of course it was Codeplex, so it was not officially supported. But with SP1, a lot of things changed in the model, so it [DAX Editor] broke.

But, I totally get it. I really sincerely hope we can come up with a better example in the product. I’m not saying that we’re doing it right now, but definitely would love to do something like that. This is part of what I was saying about having larger models. In Excel, it’s a different view point. If you are in Excel, you work to solve “a” problem, and then you throw it away. In a Tabular model, as a BI Developer, you have a solve a problem for 40 people. So, you need to look at it from all different angles, and different viewpoints. So, it’s bigger and more complex. So, you need bigger and better tools, and not just the Measure Grid.

Kasper:

One of the other examples of teams working together, and we almost had this on the Keynote: did you know you could have Excel 2013 with Power View and query Hadoop, with no caching, with our existing products today? I mean, this is awesome; it’s teams working together again. Excel 2013 Power View connects to a Tabular model in Direct Query mode. The Tabular model in Direct Query mode connects to PDW [Parallel Data Warehouse]. That sends Polybase queries directly to Hadoop. And we worked with the PDW team to make sure the queries that we send are supported in Polybase. So that they understand the queries that we send. It’s not going to be as fast as putting into Vertipaq [xVelocity]. But, there’s no caching. I directly go from your Excel spreadsheet, in Power View, to data in Hadoop and you return it.

Mark V:

How long has this been supported?

Kasper:

This has been supported for quite some time. One of my colleagues is getting in line to write a blog post about it. He still hasn’t done it. This is one of those things where, before we say anything is “supported,” we have to test it. And that costs money, right? So, that take money away from a Power BI feature or anything like that. But, in this case, we thought, “OK. This is going to be so cool!” And you can imagine, PDW just started going down this path. But, I can imagine, this will become faster in the future. So, this is going to be awesome. 

Wrapping Up

I really liked hearing how teams within Microsoft are working together. Kasper has a great point regarding traditional Microsoft BI requiring you to purchase several different products. Power BI really tosses that model on its head. if Microsoft really wants to democratize BI and bring it to the masses, the simplification of the process is a key step.

I have to confess that I had never heard of Prodiance until Kasper mentioned it. That sounds like some cool functionality that I will want to play with.

It seems that when new technologies come out, there always has to be people that say that some other technology must be dying in consequence. When Power View came out, there were people that decided Report Builder would go away. When Tabular came out, people panicked that Multidimensional must be going away. the sky is always falling, isn’t it? When Kasper made his point about the work that went into having Multidimensional Cubes support Power View, it made a lot of sense. Why would Microsoft invest time and effort in such a difficult task just to sunset Multidimensional soon after? That would make no sense. Kasper was pretty clear: Multidimensional is going to be around a while. As will MDX.

I really like Kasper’s point about Tabular being more in line with the more in tune with the Agile development cycles of today. It is a lot easier to make iterative changes to Tabular than it is in Multidimensional. At the same time, his point about Tabular not having been around the block yet is a great one. There were cool aspects to my choice of Tabular for a client project last year. There were also a few surprises that I had to deal with. I look forward to getting strong expertise with it so that I am in a better position to work around difficulties and take better advantage of new features when they come out. I was heartened by the fact that Kasper saw where I was coming from with a better environment for DAX development. Hopefully, there is more support for that within the team.

Kasper’s example of using Tabular in Direct Query mode hitting PDW is a great example of the future I would like to work in. Taking disparate technologies and putting them together to make a cool solution is just a blast.

Thanks so much to Kasper de Jonge for taking time out of his busy schedule (I think he presented 4 sessions at Summit) to sit down with me. My final interview post, with Director of Program Management for Microsoft BI Kamal Hathi, should come next week.