I missed PASS this year but I am addicted to all blog posts coming from the attendees. I am reading, sending links to colleagues and mostly thinking about the future – both with disappointment (rarely), and excitement (99% of the time). There were two very significant blog posts by Chris Webb and Teo Lachev in the last couple of days and I would like to share my thoughts on the future of SSAS.
Chris and Teo are both moderately unhappy (that’s how I got the mood from what they wrote) with the future of SSAS because of the shift of focus from “traditional” M/H/R OLAP to the in-memory Vertipaq engine, as well as the switch from MDX to DAX. My initial feeling was similar – I like MDX and the way SSAS works. Furthermore, after all, SSAS is the best OLAP tool on the market right now and it got there after a long evolution. As Donald Farmer mentioned on Tech Ed in Australia this year, SSAS is the best-selling OLAP tool for Oracle. It also is the best-selling OLAP tool in general. Leaving a tool like that and moving on to a completely new technology is a stunning move in my opinion.
There are a few things, which I think are important to consider before declaring it as a wrong move. First of all, in today’s affordable 64bit world, there are a few limits for in-memory OLAP. Also, MDX, if powerful, different, interesting and sometime elegant, is HARD. I know (personally) literally 3-4 developers in Australia who know how to write good MDX and grasp the concepts behind it. The vast majority of feedback on MDX is that it is “too hard”. Therefore, the two main selling points of SSAS – ease of building analytics and performance are hard to improve on without radical action. The only way to make Microsoft OLAP easier to use is to move on from MDX and the easiest way to improve performance is to go in-memory. This is why I am very optimistic about the future of SSAS and BISM. If small-size competitors like QlikView can build a good product (technologically), which can in some cases outperform SSAS, I am very confident that the SSAS team with its vast expertise on building the most successful OLAP tool in the world can beat them over a few years. It is better to start earlier rather than later.
On the flipside, the handling of the whole matter is puzzling and disorienting. Chris had a few really good points about the problems that may arise from such a swift transition. How do we sell traditional pre-Vertipaq SSAS when it is getting replaced with a new technology, which will require a couple (at least) releases to beat everyone else and become Enterprise-ready? This is a question I would like to get an answer for, as I am an advocate of Microsoft BI and apart from the minor inconvenience to my base for arguments with advocates of other BI suites/products will no doubt cause a major headache when talking to prospective clients. This is in fact my only concern.
To recap – I am very happy about the ongoing evolution of SSAS and I would be very happy when Vertipaq becomes the SSAS of in-memory BI, when we have an easier language to query it and when the stack gets better integration between its components. However, the transition is too abrupt for my taste. The “security through obscurity” approach to marketing I am observing may not be able to convince enterprise customers with complex needs and strategic plans to adopt a tool in transition. I hope that Microsoft does put more effort in a seamless migration path from SSAS to Vertipaq, so that the #1 spot of Microsoft OLAP technologies gets preserved through this drastic change.
Jumping straight to Amazon now, to buy a book or two on DAX – a great opportunity for me to get a head start in DAX and PowerPivot before we get the technology in SSAS J My previous concerns about the PowerPivot adoption rates because of a few little things is now replaced with excitement over the fact that we will get this new technology on a full-featured client-server architecture (if not fully-featured in Denali, then soon after that).
UPDATE: Please note Amir Netz’s response to Chris’ post in the comments – it explains the intent behind BISM and paints a brighter future for MOLAP and MDX.
UPDATE #2:TK Anand has a blog post in regards to this topic here.
UPDATE #3:Chris Webb’s follow up with a more optimistic tone here.
Boyan Penev on Microsoft BI 
Thoughtful, constructive post as usual Boyan. Echo your sentiments that the only real (and quite significant) points of concern are what to tell potential customers during the transition period and what to tell existing customers about what they should be doing with their Current cubes. Amir’s update seems to contradict the information Chris has received from the AS team. As it was with PerformancePoint, the handling of the issue is poor!
Otherwise, it’s seriously exciting times for SQL Server in general – the Denali release confirms the massive investment that MSFT is making
LikeLike
Waiting for more clarity at the moment, Amir’s comment does seem promising
LikeLike
Thinking about it more (after reading the comments by Amir and TK) and the comparisons between C++ and C#, SSAS will probably remain a solution for overwhelmingly large data volumes which cannot be fit in memory (5% of the prjects out there), while BISM-based models will probably get commonly deployed to most organisations in the future. As for an already acquired MDX skillset – well, it will probably still remain relevant to some extent, while future work will probably be directed towards setting up models and some complex calculations in BISM, as well as preparing some fixed reports in SSRS/Crescent. Not least importantly, too – if BISM drives wider adoption, we as BI professionals may find lots of new opportunities in the business sector itself…still not quite convinced by TK that MOLAP will remain as used as today, though. And Tim, yes I fully agree it was poorly handled by Microsoft in this case – with all respect, Chris Webb shouldn’t have had to break the news in this fashion and literally shock the community…Microsoft should have done this a long time ago, smoooothly, with clear perspectives, roadmap, etc…
LikeLike