On March 2nd, Ty Miller and Lothar Henkes both from SAP conducted an ASUG webinar around SAP HANA and SAP NetWeaver BW and I would like to provide a summary on this.
Please keep in mind that this does include things that are roadmap related and that these things are subject to change and that the usual SAP disclaimer applies.
- Need for a New Analytical Paradigm
- In-Memory Computing Overview
- SAP HANA – SAP High-Performance Analytic Appliance
- Question & Answer
Need for a New Analytical Paradigm
Ty explained why there is a need for a new analytical paradigm, which is based in the situation that large amounts of data have been created in each company and that it is difficult to manage these volumes. End-users are ending up waiting to get the information from the source system (often a transactional system) to their data warehouse and then finally to their BI environment. If often takes a very long time to get the critical information.
Ty then continued and explained the current options to mitigate the problem, which includes the creation of indexes, caching queries in SAP BW, creating data marts and replicating data into more reporting and analytics oriented areas, but still companies are having problems to get the right data at the right time.
Now there is the ability to store and process data in-memory and such solutions are becoming more and more affordable as the price of RAM is going down and the capacity of RAM in typical blade hardware is going up
In-Memory Computing Overview
Now in-memory computing allows to resolve some of the issues as RAM is faster than reading from disk.
Ty’s slide indicated “Technology that allows the processing of massive quantities of real time data in the main memory of the server to provide immediate results from analyses and transactions”.
The result is that the business gets immediate results from transactional data to do the analysis of data it wants to do.
Figure 1: Source: SAP
The idea of memory is shown in Figure 1 and the different levels to take advantage. HANA takes advantage of these levels of memory. Level 1 cache is a small amount of memory that can be used but it is fast. What does this mean?
This is the analogy from Ty:
If cycles were kilometers and your computer screen was the CPU …
… then Level 1 cache in memory would be the building – size of building in reference to your computer screen
… Level 2 cache would be at the edge of the city you’re currently in
… RAM would be in a different state/province/region
… Flash memory would be a different country – (or solid state memory)
… and disc storage would be the planet Mars.
For a simple application lookup, the CPU might go
4 – 8 times to RAM
2 – 4 times to Disc
That is like taking a 1000 km trip to sip your water…and a trip to Mars to refill your glass
Quote from Jim Gray, Microsoft:
“Tape is Dead,
Disk is Tape,
Flash is Disk,
RAM Locality is King”
We are switching from disk drives to memory.
SAP HANA – SAP High-Performance Analytic Appliance
HANA is a preconfigured analytic appliance. It is where SAP’s hardware partners provide the hardware and pre-install the SAP in-memory solution on that hardware.
Figure 2: What is HANA? Source: SAP
As shown in Figure 2 the hardware partners are the same as BWA solution.
Software includes tools for modeling the data – you still need to model the data but you do not need to duplicate the data. You can create the business view and made immediately available for analytics.
Data goes into HANA via a replication process – there are two ways to replicate data – using either Replication Server or Data Services for a richer ETL (extract, transform, and load) and you can allow BI on top of the data in HANA so you can have your queries, dashboards, etc.
This allows us to analyze the data that is in transactional system in near-real-time to HANA. Because the data models are already in HANA the data is now available to report on.
Some features are that the solution is fast with “sub-second update latency”, no materialized views, fast, simple and easy since it is a pre-configured appliance, with modeling based on the BusinessObjects Information Designer and SQL or MDX access.
There are two scenarios that HANA is targeted for. The first release of HANA, 1.0, is available now. The focus for release is the operational data mart – we replicating the data from an operational transactional system (ERP or CRM) system. We aren’t going through the multiple stages of an analytic environment, you replicate that data as it exists to HANA and report on those business views.
The second scenario is for an analytical data mart, so if you have a data mart or areas of a data warehouse to accelerate, you can take the data as it exists and move it directly to HANA, using Data Services or replication server, it depends on the data base that the source data resides. The data can exist in HANA and you can have analysis on that data.
Figure 3 In Memory Today, Source: SAP
Figure 3 shows In Memory today which is BWA, allowing a direct accelerated integration with BW. There is an aggregation engine that allows us to accelerate that data and stored in a column store which is maintained in memory (like HANA).
Data can come from non-SAP sources via the same Data Services tool. SAP Business Objects Explorer and the BEx client are the primary analytic tools that sit on top of BWA. The core pieces are 1) aggregation engine and 2) data is stored in a column store.
Figure 4: Source: SAP
What is new in HANA in Figure 4 is what is in the grey boxes. They have an aggregation engine, a row store, studio (a client used to model the data)
The calclulation engine is in memory as well, and this is a huge advance in processing – moving the calculation engine from the application layer but down to the database which happens to be in memory.
A row store has been added. In first release of HANA the row store is primarily used to store metadata. In future the row store will be used to run transactional database in HANA as well; it will be a traditional relational database. The column store will store analytics data.
In HANA 1.0, we have the ability to access ERP systems by replicating that data to HANA and provide SQL access and MDX.
When combining the row and column store in memory and with advances in hardware, massive throughput is possible – a “perfect storm of hardware and software”.
Figure 5: In-Memory Strategy
Figure 5 shows strategy. Today HANA 1.0 is available, replicating data from ECC to HANA and connect to existing data warehouse. It sits on BI 4.0 semantic layer to consume data with BI front end tools. As with SAP product roadmaps, the general disclaimer applies and this is roadmap is subject to change.
The next release will accelerate BW itself . The rest of the webinar focused on BW 730/BW 73x; I will save this update for the next BW 730 webinar on 4/29.
Question & Answer:
Q: Will HANA replace BWA and BW? Is that the roadmap? Why would we need BW?
HANA will not replace BW. It will eventually replace BWA and the database underlying BW. See BThe BW – HANA Relationship for more details.
Q: What happens to the BW / BWA investments already made by the user
A: SAP is developing a strategy now to ensure that investments in BWA are taken into account when adopting HANA 1.5.
Q: BW might be required for slicing and dicing?
A: This can be done when the data is coming from HANA as well.
Q: What is the use of Studio component in SAP HANA
A: To do the data modeling
Q: HANA = hardware plus software?
A: Yes it is hardware + software
Q: HANA belongs to Sybase technology?
A:It leverages Sybase technology (Replication Server)
I want to thank SAP speakers Ty Miller and Lothar Henkes for providing ASUG this webcast and Ingo Hilgefort for arranging this webcast for the ASUG BI Community.