Molson Coors talks SAP BW on HANA


While at SAP’s Business Intelligence 2014 conference, Jon Reed and I were fortunate to sit down with Pawel Moerski, Global BI specialist at Molson Coors to discuss the company’s implementation of HANA on SAP Business Warehouse. (BW)

In the ‘traditional’ SAP world, BW would likely run on an Oracle or IBM database. Replacing a database is not a trivial decision, expecially given that SAP HANA is relatively new in the market place. The question has always been whether SAP shops wold be prepared to take that step or would they simply ramp their investments in alternative databases as data requirements grow.

Molson Coors reviewed their situation and concluded that adding more Oracle licences and hardware would be on par with a HANA upgrade but they would not solve the latency problem an increased database footprint implied. I talked around alternatives such as Oracle’s Exadata and TimesTen – the possible Oracle alternatives – but for a variety of sensible reasons, these were not considered viable.

Long story short, the project was a success with three major benefits.

  1. Even though this was a database replacement, the project team found it remarkably easy to undertake with very little risk. That is unusual.
  2. Report turnaround times have been dramatically reduced so that daily reports for example are available as needed.
  3. With a little help from running HANA BW trials on the Amazon cloud, Molson Coors has confirmed that development times for new requirements will come in at least 50% less than under alternative configurations. This is because development of BW on HANA reports is much simpler than before.

In the past I along with others have accused SAP of over emphasizing the ‘speeds and feeds’ argument. After all, getting an incremental improvement in report running of itself is hardly likely to make an ROI calculation look so hot. However, it is becoming increasingly clear that with data demands accelerating, speed in development as well as turnaround processing time is the combination that customers are seeking.

In a later conversation with Holger Mueller, VP and principal analyst at Constellation Research (disclosure: I am a board advisory member) he made the point that current hardware demands cannot keep up with data demands and therefore while demand may increase, the rate of satisfaction among users will likely diminish in large workload environments as they find it harder and harder to get to the information they need.

We can have a lengthy debate about how an organization manages in those situations but if Molson Coors is indicative of what is currently achievable with like-for-like cost replacement, then SAP may well have stumbled across at least one way in which it can put off the day of diminishing returns.

Marketing criticisms aside around ‘feeds’ and speeds,’ no-one should underestimate the value this delivers to those businesses that have already invested in SAP BW and still have long life use cases and increased demand from the BW environment.

Endnote – for further discussion around this topic, check out the SCN story around the development aspects of this project.

Disclosure: SAP covered most of my T&E for attending BI2014

Correction: Holger Mueller pointed out in comments that his observation should be read in the context of in-memory costs rather than the obtuse ‘hardware demands’ reference. This is a topic to which I will return.

Den Howlett

@dahowlett

Disruptor, enterprise applications drama critic, BS cleaner, buyer champion and foodie trying to suck less every day.
Den Howlett
Den Howlett

Latest posts by Den Howlett (see all)

Leave a Reply

  • holgermu says:

    dahowlett holgermu  That was fast – no worries. I ranted on that already waayyy back in spring when it became clear that HANA would be RAM only… No need to repeat..

  • dahowlett says:

    holgermu  Sorry Holger – you’re right of course. I sense a rant coming on about this topic.

  • holgermu says:

    Nice piece as usual Den! Quick correction as it looks like we had a misunderstanding – the growth of business relevant data to be analyzed is outpacing the price drops for in memory. Luckily spinning rost gets cheaper (see Google, AWS price reductions this week) so overall enterprises who want to can keep up and store all relevant data in Hadoop clusters. Keeping it in memory only looks like a game enterprises cannot win from a cost perspective.