There was no doubt to anyone who attended Sapphire Now 2011 of the central importance of HANA to SAP’s business strategy going forward. The problem for many of us has been that the HANA information shared by SAP has been high on “game changing” ideas but not always deep in specifics. There are plenty of practical questions that need more detailed answers. In fairness to SAP, HANA development has been moving rapidly and there is a lot of ground to cover. Still, big questions on HANA remain, and you can see that on Twitter just about every day.
In fact, it was Twitter that sparked this podcast. Two fellow SAP Mentors, John Appleby and Ethan Jewett, were tweeting HANA questions and looking for some answers. The SAP Blogger Relations team proposed the idea of doing a HANA podcast with me as the moderator and referree, with John and Ethan posing questions to the HANA SAP team, and I agreed. The result was this 40 minute podcast.
As I reflect on the recording while writing up the SCN version of this podcast, I can see both it flaws and its merits. The merits? We were able to get past sound bites and into some deeper clarifications on HANA, in particular around how HANA fits into SAP’s future data warehousing strategy. I also liked that Jake Klein and Thomas Zurek of SAP were honest about what HANA can and can’t do currently. We need more of that kind of talk, clearly identifying the pros and cons of the current use cases. This doesn’t take away from the visionary stuff about HANA and what it might be able to do for the integration of OLTP and OLAP and the reduction of SAP Total Cost of Ownership (TCO) either. Both kinds of discussions are needed.
I jump in once or twice with thoughts of my own – otherwise, after my intro, this is a free-flowing discussion with SAP’s Jake Klein and Thomas Zurek fielding questions from John and Ethan. Topics such as the future of BW and data warehousing at SAP, the need for HANA benchmarking, and the reality of current HANA use cases are covered at length. Klein also explains what they are doing to bring SAP customers from Proof of Concept (POC) to Go-Live. No, we didn’t get all our HANA questions answered – time ran out on the taping first – but here’s what twe did learn.
(If for any reason the player doesn’t work, you can download the podcast using the “download media” link on the right hand side).
(Trouble downloading? if for some reason it’s not playing in its entirety for you, check out the version on JonERP.com in the meantime).
1:46 John Appleby to Jake/Thomas: What are the great use cases for HANA, and what are some use cases where HANA isn’t the most appropriate and BW and/or Sybase IQ might be better? Jake: HANA is not just an application or a capability; it’s the kernel of SAP’s next generation platform to deliver many of our applications in the future, built around in-memory technology. HANA itself is the in-memory engine, which includes in-memory database capability as well as an integrated calculation and aggregation engine.
In the 1.0 GA version, HANA is targeted for operational data mart capabilities primarily; those are scenarios where customers would like to run analytics on top of data that is sourced from their SAP or non-SAP ERP systems, and do those analytics in real time. It scales extremely well, customers can ask whatever they want without having to worry about the impact on core ERP systems databases. With HANA, they don’t have to aggregate data and do database tuning, etc for operational BI to work effectively.
So how does HANA relate to BW and to Sybase IQ and other data warehousing technology? The way SAP views BW is as an application that sits on top of the database. The plan with BW is to, over time, deliver BW on top of HANA and to selectively and progressively move and shift the performance intensive parts of BW into the HANA in memory engine starting in Q4 this year.
In its current version, HANA is a data mart, not yet a data warehouse system. It’s not designed for a large scale EDW with content lifecycle management, data management, object management and other robust EDW capabilities – those will be moved from the BW app layer into the HANA engine over time, and we’ll see the first in-memory data warehousing offering later this year: BW 7.30 that is running and optimized for HANA.
8:00 Addressing BW confusion -Ethan: what does this do to customers’ overall perception of the BW/BI architecture? How will HANA fit in? What is the end game?
12:15 Jake: One primary reason customers have built these EDW tools on top of their OLTP systems is to avoid a performance hit, so the data is replicated, transformed, and moved out. Over time, individual lines of business need to be agile and respond to changes in their business quickly. A lot of times it takes too long for IT to update this line of business information, so they end up with data marts that are geographically or line of business focus. You can have 15 EDWs and dozens if not hundreds data marts. With HANA, there is no need to move the data our of the OLTP system for performance reasons, you can simply apply the analytics tools/views on top of the operational data once that data is in-memory, and performance is not an issue. This should lead to significant data infrastructure consolidation.
15:30 John to Jake: For customers that have deployed large scale BW without the BW Accelerator – some of those data mart scenarios might be applied via HANA instead. However, there are some scenarios where HANA does provide performance problems, especially with large scale ad hoc scenarios using the BusinessObjects suite. What is the team doing to work on that? Jake: full transparency: the product is in ramp up – we have made a lot of progress in the last seven months. There are situations where customers look at HANA and see that they can deploy apps on top of an in-memory app. But the primary use case today is not to replace the database options that customers can use today; it’s also not an open database platform for all kinds of database application development. We haven’t optimized all those query scenarios yet.
18:30 John to Jake: But from what I can see, the problems occur when customers ask ad hoc questions. Other data warehousing vendors have similar issues, and they seem to be investing in map reduction algorithms to enhance performance and other ways of addressing spanning tree problems, what is SAP doing? Jake: The answer is yes. We do have algorithms to distribute queries across multiple nodes. That’s one of HANA’s strengths – to scale across deployment. For exploration and ad-hoc queries, we’re investing heavily in that area and we see that as an important area for us to excel in.
20:28 Ethan to the guys: What kinds of actions are you taking to help customers optimize their applications for HANA as you go into GA, to work with customers who aren’t part of the ramp-up program?
24:00 John to the guys: The HANA mid-term strategy looks to me like it’s away from the data warehouse. You are building out data services which will allow loading from ERP to any database within reason; you are building out HANA apps with very specific use cases. So is BW and data warehousing as a concept dead? Jake: DW is not dead from an SAP perspective at all. DW is a best practice and a set of capabilities that you need to have on top of your data layer. Our strategy is to continue to enable the best practices in the BW product today, running on top of HANA as an app with all the performance intensive aspects moved into HANA. Starting in Q4 this year, BW will be one of the apps that run on the in-memory platform as well as other apps such as Strategic Workforce Planning.
30:26 Jon to the guys: I had a reporter ask me at Sapphire Now, “Wait – are you trying to tell me that these customers aren’t live yet?” I explained many of them were in the proof of concept (POC) phase. I talked to customers about proof of concept – one was looking at HANA to apply to a high sales volume area where the BW latency issues were hurting their sales opportunities. What is SAP doing to work with customers to move them from proof of concept to go-live?
34:10 Ethan to the guys: A couple months ago, there was a release of an internal benchmark SAP did with an independent auditor to verify the benchmark findings. The idea was that the internal benchmark was very impressive. But it’s difficult to get a handle on what that means for individual customers without getting a handle on that benchmark. Is SAP issuing a recreatable benchmark that gives customers the data set and queries that were run so that it can be verified on a customer site? Jake: We’re in the process of moving the “H” benchmark through the process, before we do that we can’t openly share it or have third parties test against it. It’s something we’ve spent a lot of time defining, and it is something that is based on the TPC-H queries, however the data models and data sets have been greatly extended to reflect realistic customer use cases.
Ethan to the guys: Will it be possible to use that on another database, such as Max DB and Oracle? Jake: Once it’s approved, we would define this as a new business intelligence benchmark, and all our OEM partners would have an opportunity to executive the tests and submit the results. That’s exactly what we did internally.
38:13 Jon to the guys: Want to turn the tables and give you a chance to sound off – are there any misconceptions you’ve seen about HANA that drive you a little crazy?
Podcast links: SAP’s In-Memory Home Page on SCN. Thomas Zurek is one of the most active SAP internal voices blogging on HANA themes. You can view the Hasso Plattner and Vishal Sikka Sapphire Now keynotes, both of which were heavily centered on HANA, with a free log in. John Appleby has blogged in detail on HANA on his Bluefin Solutions blog and also on his blog on SCN. Ethan Jewett has done the same with his own blog as well as his SCN blog. I also cover the range of reaction to HANA at Sapphire Now in his Enterprise Irregular blog post: Analyzing the Real News Stories of Sapphire Now 2011, Part One: The Impact of HANA. Other links: SAP Mentor Initiative. Also recommended: SAP Mentors Vijay Vijayasankar and Vitaily Rudnytskiy have both been blogging in detail on HANA. Special thanks to Craig Cmehil of SAP Blogger Relations and the other influencer relations folks at SAP who helped make this podcast happen.