Let me continue from the point where my previous blog post “Important Questions (with Answers) about SAP’s in-memory High-performance Analytic Appliance – HANA” stops. As a reminder these are all topics that have been discussed during SAP conferences this October.
The idea of this part is to help structuring different “in-memory” products and saying what is what. The names I am using here are not always following SAP One Voice with the idea to use names we are using in our daily lingo, in some cases including as well historical names used for particular products.
5/ The world of in-memory software
First of all we need to be clear about “in-memory” in the broader context before looking at SAP specifics. Although referred as “in-memory” all products still need persistency of the data to prevent disasters in cases like power loss. The point is about the center of gravity in in-memory applications comparing to traditional.
In-memory software has been on a market for a while already. Here is my rough classification:
- In-memory BI platforms, like QlikView, Microstrategy BI or SAP BW OLAP combined with BW Accelerator.
- In-memory databases, like VoltDB, TimesTen, solidDB, or P*Time acquired by SAP in 2005.
- Databases with in-memory features, like Sybase ASE with IMDB or MySQL with main-memory storage engine in cluster configuration.
As per my previous post SAP In-memory computing engine (ICE, which is in the heart of SAP HANA 1.0) is in-memory database (category 2), but as well has built-in Calculation Engine optimized for working with SAP BusinessObjects and other products (category 1).
6/ HANA vs other SAP in-memory products
First of all, the range of existing SAP products using in-memory is broad. Among existing is not only BW Accelerator (BWA), but as well SAP APO with LiveCache, NW Enterprise Search or SAP CRM Customer Segmentation. Same way SAP High-performance Analytic Appliance (HANA) is not the only new application, but I will leave to SAP the discussion of the future roadmaps. We need to keep this big pool of different products in mind, although the focus of this blog is only on BI and Analytics applications and below is my interpretation of the road we have made in this space so far. Please do not consider this graph as a stream of the code development or software versioning, but rather an evolution of ideas and products.
BW Accelerator 7.0
For me it all started with SAP Search Engine TREX used for indexing and fast search of unstructured data, like webpages and documents. I do not know who first looked at TREX and come with the idea that it can be used for fast search and retrieval of structured data, as BW tables. For me this flash of genius truly deserved some kind of SAP’s Nobel Prize. But that’s exactly how BWA works: uses TREX to treat every record of selected BW star schema tables as a single document, that can be searched for specific attributes (like Time or Product). These “documents” are stored in “indexes” – the term that is coming from TREX legacy and is causing lots of confusion when faced for the first time by BW and Basis professionals. These indexes are then highly compressed thanks to BIA – Vertical Decomposition and put into the RAM of BWA hardware.
Because no one likes simple things, BWA itself came through multiple name changes – High-Performance Analytics (HPA), BI Accelerator (BIA, when BW 7.0 was ‘BI7’). You still can find references to HPA and BIA in the accelerator’s documentation and code. Besides the name change it has been the same version 7.0 matching the release of BW with which it worked.
BusinessObjects Explorer, accelerated
Called initially BOE by pure BW customers or BOEA most recently to differentiate from BO Enterprise, it was not new version BW own Business Explorer (BEx) as some assumed, but it was a merge of BO product called Polestar with BWA. Polestar allowed model-less exploration of the data, allowing what I call “data profiling by business users” and now has been extended to include BWA “indexes” with BW data as another source.
While BOEA gives extremely nice user experience overall and fullfills “BI for Masses” promise, it was surprising that the product targeted to BW customers had not provided integration between BO Explorer and BEx. But BOEA gave the beginning to another extremely important trend: adding in-place calculations, i.e. executing some of calculations without moving data into the application layer of BI tool and moving the technology from in-memory storage into in-memory engine.
BOEA, Wave 2
Second wave of BusinessObject Explorer (accelerated) opened BOEA to non-BW data as well. Using new piece of software called Index Designer IT can model BWA indexes and then load them from any source – not only from BW – with the special version of BO DataServices now pre-installed on the appliance.
Seen by many as a continuation of BWA line, HANA is a new product line with its own set of features.
Influenced not only by TREX/BWA, but as well by P*Time and MaxDB it has now In-memory computing engine (ICE), which is not just an in-memory storage managed using specific closed APIs. SAP ICE is in-memory database supporting access with open standards like SQL or JDBC as well as database operations like Backup&Recovery. HANA and ICE still remain SAP-centric platform. Continuing the effort initiated with BOEA, ICE will have in-place computation allowing BO 4.0 tools to delegate down some parts of processing without the need to move data between storage and application layers.
On the data acquisition side HANA 1.0 has Sybase Replication Server pre-installed that allows near real-time data replication from databases underneath SAP applications, like SAP R/3 or SAP BusinessSuite. For batch loads HANA 1.0 can receive data from BusinessObjects DataServices.
Speculations on next generations of HANA or other ICE-based applications are going beyond the scope of this post.
Often raised question is why SAP needed Sybase if they have ICE, or other way around – why ICE if there is already Sybase ASE and IQ databases available. Besides the fact that Sybase gave SAP much needed mobility platform, in database space distinction remains clear: as mentioned in previous chapter major part of ICE is SAP applications-specific in-place computation unit, which makes ICE an ideal platform to run SAP optimized products. Sybase databases remain general purpose, but as most of the other industry databases either already got (Sybase ASE) or should get (Sybase IQ) IMDB (In-Memory DataBase) features.
7/ HANA 1.0 vs BWA
In his “The BW – HANA Relationship” post Thomas Zurek already explained how HANA is positioned relatively to BW: BW covers the scenarios of Enterprise Data Warehousing and Architected Data Marts, while HANA 1.0 covers the scenarios of more departmental Operational Data Marts and Agile Data Marts.
The other often asked question is if HANA is a replacement for BWA. As the focus of this post is primarily on 1.0 release of HANA the answer is: no. For the moment BWA software stays on its own product track as a way for acceleration of BW data and acceleration of BO Explorer. The good news is that with the upgrade to SAP NetWeaver BW 7.3 you will be able to leverage in-place computation available now in BWA 7.2.
* * *
This concludes my current posts on HANA, ICE and in-memory topic, but the fact is – we are just at the beginning of the journey!
-Vitaliy, aka @Sygyzmundovych