Interesting article here on Intelligent Enterprise from Seth Grimes on how SQLStream and Mondrian can be used together for real-time OLAP analysis on Twitter feeds:
http://www.intelligententerprise.com/blog/archives/2008/12/bi_on_content_f.html
More on the technical side of this from Julian Hyde’s blog earlier this year:
http://julianhyde.blogspot.com/2008/02/streaming-sql-meets-olap.html
I’m still not convinced there’s much of a market out there for real-time OLAP (as opposed to real-time BI); after all, how many people out there are actually using pro-active caching with Analysis Services? The kind of analysis you do with an OLAP tool, looking at high-level trends in aggregated data, seems to me to be the complete opposite of BI scenarios where you need to be able to respond instantly to certain events. But maybe I’ve just not come across the right business scenarios yet.
Chris – I\’ve also been looking and working with the folks at SQLStream but come at it from a decidedly BI/OLAP background. You and I are in complete agreement that there\’s not a huge market for real time OLAP. I need to see up to the second YTD Sales to Compare with this point last YTD.As I\’ve been working with the SQLStream folks I think their technology is particularly well suited in the upkeep of the OLAP aggregates and summaries. So… I don\’t think that "up to date" OLAP makes that big of a difference in the middle of the day but think about how often you have to hit the big "update summaries/rebuild cubes" button at 2am, wait 6 hours and race against the clock. The trickle feed event processing can smooth resource use throughout the day. Instead of 10 servers in the middle of the night crunching like mad for 6 hours, the event processing engine keeps all those summaries (cube cells) up to date and then at midnight, BAM you just update them with the results. At least for Mondrian being a ROLAP tool that\’s easy enough.There\’s a lot of value in what the SQLStream types are doing but they are not in the bullseye of the traditional OLAP market. 🙂
I\’m not convinced about real-time BI either. I know PerformancePoint Server uses proactive caching, but I think that is just a work around for the performance of write back in SSAS 2005. When people ask for "real-time OLAP" I think what they really want is real time reports that run as fast as their OLAP reports.
Hi Chris,I once prospecting a customer who seems to directing BI operational in a telecommunication company. He needs to analyze and monitor all data gathering from BTS (base transceiver station) towers. I don\’t know exactly what kind of data is actually involved, but seems the analytical reports he needs has to be in near real time. Right now he claims the reports come from a 15 minutes late data.Maybe in this scenario, sqlstream seems to be fit as a solution.Regards,Feris