Editor's Note: SAPexperts' Scott Wallask recently interviewed Dr. Bjarne Berg of Comerit on his own recent HANA implementation - part of a rollout he's documenting in his upcoming HANA book with IBM's Penny Silvia. He also shares his thoughts on HANA's inevitable impact on BW and his own take on the correct pronunciation of "HANA."
Listen to the podcast, or read the edited interview here:
Scott Wallask, SAPexperts: This is Scott Wallask, Managing Editor over at SAPexperts. Thank you for taking the time to tune into our podcast today about HANA implementations.
I'm excited to once again be on the line with Dr. Bjarne Berg, Vice President of IT at Comerit. Berg, it's great to have you with us.
Dr. Bjarne Berg, Comerit: Thank you.
Scott: Now, you are co-authoring an SAP PRESS book with Penny Silvia of IBM. The book is called SAP HANA: An Introduction, which is due out by October.
What's the latest on this book, and what are book buyers going to get out of the book that they can'
t get elsewhere?
Dr. Berg: We were shooting originally for about 350 pages, but as we got into it, we realized we had to write a lot more about what you need to do and more about how to do it. So the page count kept creeping up. The book that has gone to print is basically 423 pages.
What we're doing in the book is we're spending about four chapters talking about what is the background of HANA? How does it work? Why do you care? What is the strategy? How does it transform your organization to support it?
And then we spend the other chapters talking about how to install, and we get into HANA Studio, basic and advanced concepts. And then we get Information Composer in the following chapter, and a chapter on how to load data into HANA. Then a whole chapter on how to administer HANA.
Chapter 5-11 is very much focused on how to do things, and walks you through scenarios, step-by-step, with screenshots, and explaining how things work. It's very different than some of the blogs that are out there that have a lot of great information, but very low on details.
Scott: You've implemented HANA as part of the writing of this book. Is that correct?
Dr. Berg: Yeah, we worked with IBM Labs in Research Triangle Park and they gave us a HANA system, a full box, and we actually implemented it twice. First we did a very basic install and we took it the way a hardware vendor would take it. We worked with Tag Robertson over at IBM Labs.
We basically worked step-by-step though installing it and setting it up and setting connnectivity and so forth. Then after that again, we cleaned it out and re-did it again -- this time took the standard HANA system and then loaded data to it. Then we migrated BW on top of HANA as well. We've been kicking this box around for about four m
Scott: Where do things stand now with the installation?
Dr. Berg: We're completely done. We basically finished up all the work. We did all the implementations. We did move the BW to HANA, and right now, we are actually working with some clients of ours, some large oil and gas companies, implementing HANA for real for them as well.
Scott: Reflecting back on that period of when you were installing it, what are the biggest trouble spots to prepare for in terms of time?
Dr. Berg: First of all, most of these systems are made to order. So you have to be prepared that, either you've got to talk to the vendor and see if they have an available system that they can ship fast, or you're going to have to be prepared to wait up to six weeks to get the box.
And once you get it, you have to look at the components that are installed, because there are three different versions: there are platform editions, there's an enterprise edition and an enterprise extended edition. That's the one we went with.
These have different levels of complexity and different levels of components. Some - for instance, load controller, Sybase replication servers, adaptive service enterprise (ASE) - those things only exist in the enterprise extended edition. It’s something you have to spend a little bit of time planning on -- What are you going to use it for? What components do you need? What sort of licensing? -- and then you have to size it up to make sure it's accurately sized even before you start considering getting into this.
Scott: And what have you done with HANA since you've installed it? I know you mentioned doing some work with an oil company. Are there other things you can talk about tha
t you've actually been able to do and have seen in action?
Dr. Berg: Yes. We build, of course, at the Comerit Labs. I have a lab environment with about 10 employees who work for me in a lab setting, and they have been using it extensively, both in terms of training and working.
We basically took all the two BW systems, moved them over, and did it twice. Then, we took the cubes and compressed them and converted them over to HANA optimized cubes, HANA optimized DSOs. We went through all that processing of the true migration.
And right now, I'm working with two proof of concepts with Fortune 500 companies in the defense sector. Then, we're doing the real implementation, almost at the end of it, with another one. At the same time, we have been working with some clients in mergers and acquisitions right now who have actually chosen to go the BWA route because they need to get it very quickly and before the end of the year. Then, they'll do HANA next year. So they have remote access to our system.
So there's a huge level of interest in it. Right now, I'm involved in three real projects, and of course, in the lab environment.
Scott: Have you seen any reporting, analytics or anything that you could describe, and how HANA worked at doing it?
Dr. Berg: Yes, we are using it extensively. I went out and acquired, of course, a while back a whole BOBJ suite to run part of these tests. One of the chapters talks about connectivity - how to set up with HANA. What we are seeing on the system here, when you look at the column-based data stores in HANA (versus row), in HANA column-based, we are seeing 7, 8, 10 times faster response times, even relative to BWA. When we compare it with our traditional BWA system, which is actually fairly small -- it's about 700 or 800 gigs, so close to a terabyte -- in that one, relative, when we
turn off cache, we are seeing performance turning on the very large data fetches in the neighborhood of 2, 2 ½ thousand times faster.
So we're talking queries that we deliberately made slow by accessing a lot of data, a lot of calculated restricted key figures. Those queries were running at about 30, 35 minutes, and are now going in a few seconds.
Scott: Wow! That's incredible.
Dr. Berg: Yeah, this is really a gateway technology that is not just making things faster. The major thing, that I think a lot of people don't understand yet, is that it also replaces the database and the application server. So when you look at the application server we use in BW today, that also becomes in-memory.
What people don't realize is that things that we do in the BI analytical engine today, which is calculated key figures, restricted key figures, where we're doing sorts, we're doing restrictions on hierarchies, all the things that happen in the BI analytical engine, it also happens in memory. It's much more than just grabbing data and presenting it really fast. It's doing all the calculations level as well, and that's where it gets fantastic. There's no data transfer between servers here. It goes straight from memory and reads and calculates, and that's when we see spectacular performance increases.
Scott: We did a Q&A chat with you and Penny Silvia back in July. During that chat, you discussed migrating BW over to HANA. Is that a standard type of migration, or does it present any unique challenges?
Dr. Berg: Actually, there are some migration kits out there. You can move the migration and the stuff over to BW. The only thing you've got to consider here is there's a few more features inside HANA. We talked about being able to optimize the infocube, which physically changes the da
ta structure of the cube. It makes it much faster, and gets rid of the dimensions. You can also do optimizations of the DSOs. And the way we do compressions also changes inside of it.
So when we say this, we tell our clients, "Don't just take HANA and install it and slap BW on it." You probably need to go a little deeper and pay particular consideration to non-cumulative key figures, like in inventory cubes.
So it's not a slam dunk, but it's actually fairly straightforward, if you have the right resources.
Scott: There's been a lot of discussion over the last few months online about whether SAP is setting up HANA to replace NetWeaver BW, and what that means.
Where do you weigh in on that whole debate of HANA replacing BW?
Dr. Berg: It's kind of philosophical here.
I think it absolutely replaces BW from a conceptual standpoint. From a physical standpoint, we will not be using BW the way we've done it -- I've done it -- in the last 13 years.
It was all based on relational databases, getting the power of ROLAP, Relational Online Analytical Processing, and getting the strength or the speed out of the relational database, using dimensional data modeling and star schemas, and so forth.
I think the era of that is going by the wayside. I don't think we'll be doing that in the very, very near future.
Having said that, I still think we need to have a BW-like -- whether we call it HANA or we call it BW - we still need to have layers of reporting, where data is integrated, different granularity, different analytical purposes, different calculations, different access levels.
So I think we still need the data warehouse to be a repository for the data, but I really doubt it's going to be back in relational databases. Within three to five years, I expect everybody to be on a HA
NA system. There will always be some stragglers, but those will be an extreme, small minority.
So, yes, BW is going away, in the way we're using it today, but it's morphing into something that's substantially better, and substantially faster.
Scott: And finally, Berg, we here at Insider Learning Network and SAPexperts are running an international poll, which is completely unscientific, about how to pronounce HANA correctly. And I want to share the current results with you which indicates 65.2 percent say “hahna”, and 34.8 percent say “hanna”. So, do you have any reactions to this breaking news?
Dr. Berg: I thought it was just my wife from the South. She would probably say “ha-na”. But, coming from Europe, I'd probably say “hah-na”.
Scott: Me too. It seems like we may have to do some retraining on the SAP end of it, how to pronounce the product correctly!
On that note, Berg, as always, thank you for your thoughts and advice for our listeners out there.
Dr. Berg: You're welcome.