Having spent a few days in Mid-January 2012 getting “Hands On” with SAP Hana and interacting with it in the BusinessObjects BI4 toolset I thought I’d share some initial thoughts and for a change I’ll start with the summary first.
The value that I see Hana bringing to organisations is not just faster response times to questions (BI Queries) but the business value of new insights that can come from asking more and different questions from larger data sets than possible before. Additionally there are significant business opportunities for companies by bringing to new offerings to market that have not been possible thus far, for example by tracking behaviour and facets in real time and taking action to either up sell/ cross sell to customers or byoptimising the supply chain immediately following world events for example recent flooding in the Philippines.
I am hearing many examples recently of companies where the analysis of their disparate structured data (let alone unstructured) is soo difficult that end users just give up trying and say it is not possible. If you have to wait hours for a query over “Big Data” volumes to return would you really keep driving the questioning to find the nugget in the bottom of the Gold Pan? Whereas, by bringing analyses that were not possible before into the hands of users of all levels in seconds not minutes or hours imagine the new opportunities and efficencies this could bring.
However, an area where I am still pondering is that SAP has not one but two offerings in this database space, SAP Hana and Sybase IQ. The entry costs for companies to start the journey into Hana is on the face of it high, I see the return on investment argument, but the price tag for SAP Hana could be challenging for the SME market place. Is this where Sybase IQ fit’s in the picture ? On the face of it IQ is a columnar data base just like Hana but the data persists on disks rather than in memory. Experience suggestd that you would expect database performance and the end user experience to speed up significantly using IQ without the need for proprietary hardware from a selection on vendors. If IQ maintains a lower entry cost for the SME market place, should we consider the 80/20 rule .. Food for thought and discussion I hope.
Turning now to my hands on experience, my first impression getting hands on with the Hana Studio application is that it looks and feels very similar to universe design as the BOBJ classic audience would know it, simple and effective. However, a significant and understated benefit of the Hana Studio design approach along with the lack of data physically persisting is the ability to rapidly change, augment and bring on line new windows into “Big Data”. A change that would take weeks or even months tom deliver in a multi-tier BW landscape could be implemented in a matter of days, just imagine how that alone could transform a business……
Finally, the topic of sizing a Hana server is an emotive one in my short experience thus far. A Very Very basic sizing guide for a Hana server could be:
- Take the size of the ERP base table in GB and divide by 5 (compression ratio), then double the number as headroom need for internal Hana operational activities.
- Example – 500GB base data table /5 = 100GB then double it to 200GB to allow for headroom.
- So 500GB of data will require a 200GB in memory element for the Hana server.
- It has been suggest the compression ratio can be up to a factor of 20 in marketing collateral, I am being cautious in the calculation above