PASS Summit Interview With Kamal Hathi

13 November, 2013 (10:00) | Azure, Interviews, PASS, SSAS | By: Mark V

For the third, and final, installment in my PASS Summit Interview series, I present my interview with Kamal Hathi, Director of Program Management for Business Intelligence Tools at Microsoft. Kamal is the one who is ultimately responsible for the direction of Microsoft BI.

As with my other interviews, the byproducts of casual conversation have been edited out for better flow in writing.

Transcript

Mark V:

Things are changing very rapidly in the BI space. There are so many tools coming out and Microsoft is delivering so many awesome new features. As a BI Developer, what does my job look like 5 years down the road?

Kamal:

That’s a great question. I think we’ve been on a journey where BI was something central that somebody built and everybody sort of just clicked and used, or maybe had a report. Now, we’re coming to a point where a lot of it is empowered by the business. The business guy can go off and do lots of things. But there’s two parts that are interesting that a Developer, a Professional, if you will, needs to be in the middle of. One, I think, is the data. Where does the data come from? The first requirement is having the right data. And call it what you want, Data Stewardship, Data Sanctioning. And it’s not just Data meaning rows and columns or key-value pair kinds of things. I think there’s Data, as in the Data Model; building a Data Model. And a Data Model, sometimes, is not different than what people do building Cubes. They have Hierarchies, and they have Calculations, and they have interesting drill paths. All kinds of things. So, someone has to go do that work, many times. Even though, I think, end users can go and mash up data themselves. But there are times when you need a more supervisory nature of work; or a more complicated nature, if it turns out that Calculations are difficult, or whatever. Someone’s going to always do THAT work. And that’s interesting. The second piece that is Interesting is that I think there’s going to be a new work flow. We’re already seeing this pattern in many places. The workflow is going to be like this. An End User decides they want to build a solution. And they go off and, essentially, put something together that’s very close to what they want. But it’s not really high performance; it’s not perfect. Maybe it’s got some missing things. And they use it. And their peers use it. It gets popular. And then some Developer comes in and says, “Let me take that over.” And then they improve it, make it better, polish it up. Maybe completely re-write it. But they have the right requirements built in right there. A third piece you will start seeing is going to be the Cloud services based world. That is taking bits and pieces of finished Services, and composing then together to build solutions. And that’s something we don’t see much today. But I can imagine someone saying, “Hey, I’ve got this Power BI piece which gives my visualization, or some way to interact. I can take some part of it and plug it in to another solution. “ They can provide a vertical solution, provide a customized thing for whatever audience it is. And be able to do so. I think those kinds of things will be much more likely than doing the whole thing.

Mark V:

So, instead of necessarily the BI Developer working with the users to help map out requirements, with these End User tools, they can build something that’s their requirement.

Kamal:

More or less. And then the Developer can jump in and make it better. Maybe re-write better expressions and queries and make it faster; all kinds of interesting things. It just adds more value to their job. Instead of sitting there talking to people and getting “Oh, this is wrong. Re-do it.”

Mark V:

When Tabular came out, there was some uproar. When a new feature comes out, there are always “the sky is falling” people that say that something else must be going away. It happened when Power View came out. People said, “Report Builder is going away.” Then, when Tabular came out, “Oh! Multidimensional is going away.” I still hear it out there sometimes that Multidimensional is dead or dying; MDX is dead or dying. What message do you have for those people?

Kamal:

Two things here. MDX and Multidimensional are not the same thing. We should be very careful. Multidimensional is very important. Lots of customers use it. And we’re actually making improvements to it. This latest thing, DAX over MD, which allowed Power View to work over Multidimensional, is a great example. We know this is important. There are many customers who have very large scale systems in Multidimensional. And it’s important to us. We’ve just come to a point with Multidimensional where large jumps in functionality are just harder to do. You can really go on an say we can 10X anything. And so the In-memory stuff, the Tabular, has been the approach we’ve taken to give more kinds of scenarios; the more malleable, flexible stories, the “no schema design up front” kind of approach. But Multidimensional is super important. It isn’t going anywhere. So, when someone asks, “What should I use?” we say, “You pick.” Our tools, our aim is, should work on top of either. We would like Tabular to be on parity with Multidimensional in terms of capabilities. And we’re making progress. We’re getting there. We haven’t quite gotten there. But uses shouldn’t have to worry, in our opinion, about Multidimensional or Tabular. These should be things that you worry about as a tuning parameter, but you shouldn’t have to worry about them as a big choice.

Mark V:

So, the Business shouldn’t be concerned about it.

Kamal:

Right. It’s a store. It’s a model. It’s a calculation engine. And you might want to switch one to the other. And in the future, that could be a possibility. There are lots of limitations that might not make that practically possible, but notionally speaking. I’m not saying you could just pull a switch. But you should be able to have similar capabilities and then decide which one to use.

Mark V:

Or possibly some kind of migration tool?

Kamal:

Yeah, but those kinds of things are harder sometimes. They’re not so easy to do because.. who knows what’s involved? What kind of calculations did you write? Etc. Those are much harder to do. Migration is always hard. But comparable capabilities make a lot more sense. So, I can take this guy, and build the same thing and not have to worry about a dead end.

Mark V:

When I was at TechEd, Kay Unkroth did a great Managed BI session. And he started with a demo of taking in data from the Internet and combining it with some business logic. It was tracking the purchasing habits of Pink Panthers buying cell phones. And in his scenario, they went through a big investment at this Company only to find out that Pink Panthers don’t exist. So, in the realm we have, with data becoming more self-serve, with Power Query, etc, being able to reach out to more and more places, what is the thinking [from Microsoft] on the ways we can continue to have some governance over the realm that users have?

Kamal:

Fantastic. This question goes back to the first question on what Developers do. We talked about Data Stewards and Sanctioned Data and all that. And even with Power Query, if you work with that, the catalog you find things from isn’t just the Internet. The Internet is one catalog. We are also enabling a Corporate, internal catalog, which today, we showed in the keynote. And you saw what we can do. It’s in Power BI, and you can actually try it out. And the goal there is to find a way for someone, it could be a business user, maybe a Developer, the IT professional, to go off and add to a catalog that is internal to the company. They can add something they think is trustworthy and worth sharing. And then someone else can come in. And they want the shopping habits of certain constituencies or a certain segment, they can find it. As opposed to “I heard on YouTube that 20-inch phones are now hot.” Who knows, right? Just because it’s on the Internet, doesn’t mean it’s true. That’s the idea of having this info catalog, essentially. It knows how to provide people the capability of publishing. And that can be a mechanism for whoever one deems fit in the process to provide that sanctioned publishing. And maybe have a data curator look at things and make sure users have a place they can go for trusted sources as opposed to “wide open.” And we actually enable that. People can do that.

Mark V:

Could you then disable the Internet piece of that catalog in certain organizations?

Kamal:

Potentially. The question is Why? And, again, it’s a process thing. You ask people Why and what they want to do. And that’s the danger of governance. The minute you make something such that it becomes that Forbidden Fruit, someone will find a way of doing it. They’ll go out and Cut and Paste. They’ll do something. It’s almost impossible to really lock these things down. What you do want is people to be aware that there’s an alternate, a sanctioned place, where they can go compare and they can understand. And that’s what they should be doing. I think it’s much harder to lock it down and say you can only get data from inside. But that’s a process view; and organization will decide how they want to do it.

Mark V:

Early in my career, I built several report models for clients. Obviously, you know, they’re going away. They’re not be invested in. I recently wrote a Technical Article about alternatives to Report Models in SQL 2012. And I laid out the different options that you have, from Power Pivot for Excel, Power Pivot for SharePoint, Tabular, or full on Multidimensional, etc. One of the things that was pointed out to me after that one of the things that goes away with report models is the ability to make that regular detail style table report in Reporting Services that can go against a drag and drop model that’s not a Pivot table. Is there something coming down the road that could alleviate that?

Kamal:

There are ranges of this, right? You can do a drag and drop tablix, even, in Power View. So, it’s not like it’s completely gone.

Mark V:

Right. It’s not completely eliminated. But it’s not the flexibly of a Report that you can do subscriptions with, etc.

Kamal:

I think in terms of those kinds of things, like Subscriptions, mailings… Two things. One. Reporting Services is still there. It’s not like it’s gone away. And number two, we obviously think it’s important to have the same capabilities in other places and so we’re looking into making that possible. A lot of people ask for it. And it seems like a reasonable thing to ask for. Why would I not want to go schedule even my Excel sheets? A report’s a report. For some people a report is a Reporting Services report. For some people a report’s an Excel report. It’s an important point and certainly we’re looking into that as something we would do. But I don’t know When, Why, How, Where. As with all things Microsoft, it’s open to interpretation by whichever tea leaf you read about.

Mark V:

From the visibility you have within Microsoft to see what’s going on, could you describe, in a general sense, how a feature for, say, Analysis Services, goes from ideation, with someone saying, “Hey, wouldn’t this be cool?” to making it into the product?

Kamal:

Absolutely. There are multiple ways this happens. One is when we had an idea before or a customer had asked for it, and we decided to ship it and couldn’t. And then it becomes a “backlog” feature. So, next time, when the next version comes, this one is left over and we say, “Hey, come join the train.” We just pull them back on. That’s easy. Done. Everyone knew we wanted, or tons of people had asked for it, and we just couldn’t finish it, so we just put it over. The second, which is much more likely, is customers have been asking for something. For example, we heard the feedback loud and clear, “We need support for Power View of Multidimensional models.” That’s a pretty clear ask. The work is not easy, but it’s a pretty clear ask. So then you go and say, “What does it take?” We figure out the details, the design, the technical part of it. And then you figure out the release timeframe, the testing. All that stuff. And you do it. The third one is when somebody say, “I have an idea.” For example, “I want to do something. I have an idea for an in-memory engine and it could be like a Power Pivot.” And that’s just like, “Wow! Where’d that come from?” And then you say, “Will it work?” And nobody knows, right? So then we start going in an asking customers what they think. Or, more likely, that idea had come from talking to somebody. And this Power Pivot case actually came from talking to customers that said, “Hey, we want end users to be more empowered to do analysis themselves. What can you do?” That was essentially the germination of that idea. When that happens, there’s usually some customer input, some personal input. And then they start to come closer to fleshing it out and ask who’s going to use it. So we ask customers, get some replies. Sometimes we just do a design on paper and say, “Would you use this?” And then people try it out, give it some feedback, and say, “Yeah. That’s good.” So, we go down this path getting more and more input. And then we come up with a real product around it. Sometimes, we have an idea that we think is great and we float it around and we hear vocal feedback saying that’s a bad idea. And we go back to the drawing board and re-do it. And this happens all the time. Many of these features show up, and people don’t realize that we went back and re-did it because the community told us. And many people ask if we use focus groups here [at PASS Summit]. There’s one tomorrow, actually, to go sit down and say, “What do you guys think about XYZ?” There’s feedback; and we listen. So, there’s an idea, there’s feedback, there’s evolution. And we iterate over it all the time until we come to a solution on that. Very rarely to we just build something where we didn’t ask anyone and we just did it. We come up with a brainstorm and flesh it out, but we always get feedback.

Mark V:

So, Matt Masson has discussed the debate within the SSIS team regarding the Single Package deployment model versus the new Project Deployment model. I would assume that happens throughout all the products. Maybe there’s somebody championing this feature and people asking “What are we going to do?”

Kamal:

Yeah. My first memory of working on this team is people shouting in hallways… in a friendly way. People shouting, “Hey! What about THIS?” or “What about THAT?” And it turns out to be a very vocal and a very passionate environment. People don’t just work because they walk in in the morning and write code. They work on this because they are just in love with this product. They are just deeply, deeply involved and committed to what they’re developing. And these people just have opinions. And I don’t mean opinions like “Yeah, I want to have a hamburger.” They have an OPINION. And there are very passionate discussions that go back and forth and back and forth sometimes. Typically what happens is a Senior Architect or someone who has some tie-breaking capability can come in and say, “Look, great ideas everybody, but we’re going to do THIS.” And then people listen, there’s some more argument, and after that, it’s done. And we go and do it. And Power Pivot is a great example of that. People were like, “Are you crazy? What are you doing??” And it was like, “No. We’re going to do it.” And that was it. And the team just rallied behind it and built a great product, and off we go. But, the good part about that story is that, because people have such vocal opinions, they rarely remain silent about these things. We don’t lose anything. And so we have an answer that we can listen to. And we can decide from multiple options as opposed to just going down one path. And then we end up with a decent, rigorously debated solution.

Mark V:

So, there’s a lot going on right now with Cloud and Hybrid solutions; there’s Chuck Heinzelman’s great white paper on building BI solutions in the Cloud in Azure virtual machines. When I talk about that with clients or with colleagues, there’s still a lot of trepidation and people say, “For some industries, that will never ever happen.” What kind of message would you have for people that just don’t see how it’s feasible?

Kamal:

Cloud has many aspects to it. There is the “Everything’s in the Cloud. I want to put my financial, regulatory data in the cloud.” Which, is not going to happen for some people. Then there is the “I can get value form the Cloud on a temporary or on-demand basis. I want to go burst capacity. I want to go off and do a backup.” Whatever it is. There’s that. Then there is the “I’m going to do some parts of my workload in the cloud and the rest will remain on premise.” And for all these options, whether it’s completely in, hybrid, temporary bursting, that value provided is typically what customers decided makes sense for them. If it’s the value of all the infrastructure just being taken care of and I can do more things to add value to a solution, so be it. If it happens to be that I can get extra capacity when I need it, great. If it happens that I can keep my mission critical or compliance related data on premise and lock it up, and hybridly work with the cloud, that also works. And for most customers, most users, there is value in many of these scenarios. And there isn’t any one answer. You can just say that “The Cloud” means that you do X. it just means that you have many options. And interestingly, Microsoft provides many options. We have Platform [PAAS], fully deployed platforms that you pretty much just deploy your database and have to do nothing. Azure SQL Database, good example. All you do is worry about building your database, setting your tables, and you’re done. We take care of all kinds of things in the background. Don’t like that? Go to a VM. And you can do all kinds of fancy things. Chuck’s paper is a great example. We have solutions. Office 365 gives you end-to-end; from data to interfaces to management, all in one. Each of these things have hybrid solutions. You can have data on premise. As we saw today in the keynote, you can backup to Azure. With 365 and Power BI, you can actually do hybrid data connectivity. So, all of these things work in different ways. I think, for every customer out there, it typically is just a trial solution, they want show something, or they want to have part of their solution that works. Either way, it adds value. Typically, what I have found as I talk to customers, is that many of them come from the “I would never do that” to “Oh. I might try that.” Because they have come to the realization that it’s not an all or nothing proposition. And you can do parts, try it out, and it works.

Mark V:

Over the Summer, I did a proof of concept for a client using Tabular. Just because, with the size of it, and looking at its characteristics, I said, “This would be a great Tabular solution. Let me demo it for you.” I have talked to several people in the industry. And the process of developing Tabular in SQL Server Data Tools can be… less than awesome. I had some bumps in the road. There was a great add-in I got from CodePlex [DAX Editor] that helped me deal with DAX in more of a Visual Studio environment. It didn’t apply well to Service Pack 1 [of SSAS 2012 Tabular] and that kind of stuff. There was something that Marco Russo had put forth on Connect that suggested more or less a DDL language for Tabular models. Is something like that feasible?

Kamal:

I don’t know. I don’t have a good answer for that. That reason for that is we’re looking at many options. What would move the ball forward in that direction for a design environment for Tabular models? And there are many options. So would say let’s do it in Excel; take Power Pivot and put it on steroids. It’s a possibility. Or a DDL language. Go off and take the things you had in MOLAP and apply it here, maybe. Maybe something brand new. I don’t know. We’re trying to figure out what that is. I do know that we do want to take care of people who do large scale models, complex models in this environment. I just don’t know how and where and when. But it’s an important constituency, an important set of customers. And we’ll figure out how to do it.

Mark V:

As a BI developer, it’s important to know that the discussion’s being had.

Kamal:

All the time. This happens in a lot of hallway discussions. This is one of those.

Wrapping Up

There’s a little bit of a story to this one. When I decided I wanted to do a blog series composed of interviews conducted with Microsoft folks at the Summit, I wanted to get different perspectives. With Matt Masson (Blog|Twitter) and Kasper de Jonge (Blog|Twitter), I already had members of teams in the trenches of development of the tools. I had then reached out to the awesome Cindy Gross (Blog|Twitter) to get the perspective of someone on the CAT (Customer Advisory Team). Cindy got back to me with a contact for Microsoft PR, Erin Olson, saying that she was told to send me there. Upon contacting Erin, she responded by offering to have me sit down with Kamal Hathi, who would already be on site that day. That was an offer I couldn’t refuse. In hindsight, I am wishing I had asked about sitting down with Cindy as well, but I had already decided that my first series of this sort would be capped at 3 since I had never attempted anything like this before and didn’t know what to expect. If this series proves to be popular and of value to the Community, then I will certainly consider doing it again and asking Cindy to participate.

You will notice some overlap in the questions posed to my fantastic interviewees, particularly between Kasper and Kamal. I wanted to get different perspectives from within Microsoft on some similar topics. I also made sure to branch out in each interview and ask some questions targeted to a particular person.

In response to my “5 years down the road question,” Kamal echoed the importance of Data Stewardship. It is clear that this is an area that Microsoft is taking very seriously. Having done a lot of reporting in my career, my motto has always been, “It HAS to be right.” Clients have appreciated that. As we open up more and more avenues for users to get data, we must keep in mind that the data needs to be trustworthy. 

I really want to highlight the ways in which Kamal described how a feature makes it into the product. Make special note of the fact that customer feedback is vitally important to Microsoft. Sometimes, the idea itself comes from Customers. I think Microsoft often gets a bad wrap as some kind of bully or something merely because it is big. It is certainly not perfect; no company is. But I think it is really important to make note of how Microsoft DOES listen to customer feedback when it comes to the products they provide.

Kamal’s description of the internal debates that occur within Microsoft over features is important. It also echoes what we heard from Matt and Kasper. The people working on these products for us care VERY deeply about what they are doing. The work and passion that go into creating these tools we use every day is staggering to me. While I have never been a “fan boy” of any company, I have chosen the SQL Server related technologies upon which to base my career. And I have no regrets. This is a hugely exciting time to be a BI professional. The investments that Microsoft have been making in this space over the past several years make it even better.

This concludes my PASS Summit Interview series. Thanks so much to Matt Masson, Kasper de Jonge, and Kamal Hathi for taking time out of their very busy schedules to sit down with me and answer my questions. Thanks also to Cindy Gross and Erin Olson for their assistance in connecting me with Kamal. This series turned out even better than I had ever expected thanks to the generosity of those involved.

Comments

Pingback from Looking Forward to the PASS Business Analytics Conference | Mark V SQL
Time May 1, 2014 at 10:00 am

[…] of Microsoft. These gentlemen are both brilliant and great speakers. I actually had the honor to sit down with Kamal for an interview at the 2013 PASS Summit in Charlotte, NC. That was a great […]

Write a comment