Last month, Uddhav Gupta, SAP, provided an ASUG webcast to the High Tech SIG. As always, the usual disclaimer applies that things are subject to change. Below is a summary of what I learned.
The purpose of this webcast was to discuss what the product is all about and the roadmap going forward. What is in-memory computing and what is the context? For SAP, it is processing massive quantities of real-time data in main memory, without building aggregates. The idea is to provide immediate results using the BI clients.
Figure 1 – In-Memory Computing Strategy (Source: SAP)
Figure 1 shows the guiding principles. It is at the heart of future applications with co-innovation projects with customers. Hardware partners deliver the hardware for the product.
Figure 2 – Source: SAP
As Figure 2 shows, in-memory solutions are not new to SAP. The BWA is the “star” product, designed to accelerate info cubes and queries. BusinessObjects Explorer is on top of BWA for an accelerated solution. Enterprise Search is based TREX. Business ByDesign uses in-memory technology.
Figure 3 – Source: SAP
As shown in Figure 3, today a single blade you can have 64 cores, with 64 bit address space you can address up to 2 terabytes of main memory on a blade. On the software side, the entire row store can be in memory; SAP is one of the first vendors to have this. Insert only on Delta is fast, and you can always go back in a period of time to see what the results looked like.
Figure 4 Using Columnar Data Store – Source: SAP
Figure 4 shows how the conceptual view maps in memory to the row store and column store memory addresses. Column store in memory can pick up chunks of data.
Figure 5 – What is HANA? Source: SAP
HANA is a preconfigured appliance, similar to what you buy with BWA. You buy the hardware from the hardware vendor, buy the software from SAP, and the hardware will come pre-loaded with the software. HP, IBM, Fujitsu and Cisco are the hardware vendors.
Figure 5 on the left shows 5 main building blocks. The first one is the in-memory computing engine which is a full fledge in-memory row and column store. Then there is the calculation and planning engine, which has the OLAP capabilities in it. To bring in data to HANA there are two mechanisms. One is real time replication services from ERP system into HANA to keep in synch with ERP. For others you can use Data Services from SAP and non-SAP systems into HANA.
Once data is in HANA, you use the Modeling Studio, based in Eclipse. The modeling studio allows you to administer HANA, model data, build calculation views, analysis authorizations, and analytic views within that one single studio.
On top of HANA you have the BI Consumer Services interface, SQL interface, and the MDX interface which will allow the BusinessObjects clients, as well as other applications to access data within HANA. Excel is the only non-SAP supported application to access HANA.
Figure 6 – Example – Source: SAP
Utilities are moving to smart meters, collecting data from meters every 15 minutes. This helps determine how much power to buy from the grid. This provides real time analysis on huge volumes of data.
Figure 7: Retail Example Source: SAP
Analyze all point of sale data to predict demand in a timely manner. One customer for example is using 40 TB data in memory using compression to fit into memory and using BusinessObjects Explorer to report on it.
Figure 8 – Source: SAP
HANA has great power when it is used as underlying engine for your applications. Using Business ByDesign on HANA, they put the SAP Dunning process, with 154K business partners, 70 dunning items per partner, 1.8 million records were read. In the standards system the dunning process took 77 minutes whereas it only took 13 seconds in HANA.
Figure 9 – Roadmap – Source: SAP
Figure 9 shows HANA 1.0 is in ramp-up today. Capabilities now are providing real time operational analytics, replicating the ERP system into HANA, creating calculation and analytic views to report using BI tools out of HANA. BI 4 will run on HANA. The benefit is real time analysis of operational data.
In the mid-term, which is the later half 2011, early 2012 SAP will have next generation applications. SAP BW will run on HANA and enable BW on HANA. The SAP Business Suite will run on HANA. In the long-term, which is 2012 to 2014 there will be the persistence layer on the Business Suite.
Figure 10 Source: SAP
On the left in Figure 10, today we use the ABAP Application Server. This ABAP application server performs the reads and writes to database and this takes up time.
The next generation applications will delegate to in-memory platform like BW does to BWA. For example, BW delegates OLAP functionality to BWA such as top N, currency conversions – these processes occur on BWA side before giving back to BW .
Figure 11: BI and HANA Source: SAP
How do different BI clients leverage HANA 1.0? On the bottom of Figure 11, you have BW and ERP. ERP is replicated to HANA, either in its entirety or select replication. BW is loaded to HANA using Data Services.
Crystal, BusinessObjects Explorer, Analysis Office will directly use the calculation and analytic views in HANA. WebI, Dashboard and Crystal will go through the semantic layer to access HANA. Microsost Excel will use MDX ODBO interfaces.
Figure 12 Source: SAP
As shown in Figure 12, today we have ERP data, data marts, and non-SAP data in a data warehouse such as BW.
Figure 13 Ramp-up Today – Source: SAP
In Figure 13, data is replicated into HANA real-time. This is called the Operational Data Mart scenario. Also, data is brought into the BW system, still using BWA to accelerate queries.
In the analytic scenario data is sent to HANA real-time, building the calculation views and expose out to local BI clients. It is fast as data is in memory.
Figure 14: Mid Term Plan Source: SAP
In the mid-term plan in Figure 14, BW will use HANA as a persistent layer. HANA will have Virtual Data Marts allowing replication of data in real time. Non-SAP can use ETL from Data Services into HANA.
Figure 15 Mid Term – Source: SAP
Figure 15 will feature the Next Generation ABAP, Java and embedding HANA.
Applications layer will be capable of leveraging HANA with new applications on top of HANA.
Figure 16 Longer Term Plan Source: SAP
SAP BW and ERP run on top of HANA. OLAP and OLTP are in single persistence layer. Note that the ETL between ERP and BW are happening in same layer in real-time.
At this point, Uddhav provided a demonstration of the HANA tool.
Figure 17: Information Modeler Source: SAP
Figure 17 shows the Information Modeler with its functions.
Figure 18: Administrative Mode Source: SAP
Figure 18 shows the administrator mode showing the HANA console, providing an overview of the system. It shows how much data is being used.
Figure 19: Tables Source: SAP
Figure 19 shows the tables in the catalog. This was extracted from BW, showing the DIMS and the FACT tables.
Figure 20 – FACT table Source: SAP
Figure 20 shows the FACT table.
Figure 21 – Information Models Source: SAP
Figure 21 shows the most important part of HANA – the information models, with the VEHICLE_SALES_ACTUAL as the Analytic view.
Figure 22 Logical View of Data Foundation Layer Source: SAP
Figure 22 shows the logical view of the Data Foundation layer.
Figure 23 Security Set-up Source: SAP
You assign the roles, users, and you can also import the roles.
After this, you can expose the analytic view to BusinessObjects Explorer.
Figure 24: BusinessObjects Explorer using HANA Source: SAP
In Manage Spaces, you see the In-Memory Computing Engine, which is HANA and the Analytic View. Then you can use the standard BusinessObjects Explorer to edit the facets, measures, etc.
Figure 25: Exploration Views Source: SAP
Then you can go into Exploration views to slice and dice the data however you want.
He ran a query against 61 million records in .623 seconds (not indexed).
It was great to see a HANA demo, especially on a webcast and see how it works. We saw the Information Modeler and how a BI client such as BusinessObjects explorer connects to HANA. It was great to finally see the HANA screens.
I want to thank the ASUG High-Tech SIG and Uddhav Gupta of SAP for a great webcast.