(responding to a recent post in c.d.o)
God it's good to know that there are still some Oracle people out there who can cite chapter and verse. I was beginning to think you guys had dropped off the face of the planet.
Your explanation for the death of the APB-1 is perfectly acceptable to me, and that was the source of my information, plus one other non-disclosed perf comparison between Essbase and MSAS dated last summer.
So if a new benchmark can arise and a standard way of describing systems that is not so dependent on Nigel and us biased old heads, that's all good. As an architect with over 17 years in the business, I would have loved to learn as many products as possible, but it hasn't been possible in OLAP because of this lack of interoperations.
I am hopeful that era is coming to a close as premier vendors such as Hyperion, and hopefully Oracle can implement things like external authentication to LDAP etc and deeper integration to UML driven design tools that make enitre infrastructures more clear. Again, nobody should be stuck on an entirely proprietary platform.
As for Hyperion and performance, what I expect to see is that their partnership with Teradata will bear fruit. That may mean, as a standalone OLAP engine, Oracle OLAP may handle the higher volume apps a bit earlier withing MOLAP storage. But since Hyperion already has Hybrid (which is pure relational storage mapped under the OLAP access methods and calculations) large cardinality dimensions are handled nicely. At a certain point, with Hybrid, Essbase could leverage Oracle RDBMS as well as OracleOLAP. I think of it as a slider. Where do you want to store what part of the big model?
It's true that Essbase Aggregate Storage is lacking in the full set of Essbase calculation functions but it is demonstrably very well capable of those applications which were fairly well understood at the time to be pure ROLAP apps, like retail. By implementing this multi-kernel strategy under the same API, Hyperion has a done a very good job of expanding the application envelope for the Essbase platform. ASO is a killer on performance and 50 times performance over Essbase 6.5 is probably typical and 400x is not uncommon. I still don't understand the underlying technology. To me it still seems like magic. Obviously I'm looking forward to some bake-offs.
As I implied earlier. The scalability issue is almost nullified. From a Hyperion perspective, it's a short matter of time before they integrate ASO with the fuller set of Essbase calculations (and user defined java extensible custom algorithms) plus add support for slowly changing dims. 18-24 months at the outside is my guess. So we will have a two order of magnitude jump in performance. Add to that 64bit, which is already available which bumps everything about 1.4 with no tuning, and you're talking about imagination as the limit for apps. People are going to have to start thinking of new classes of analytical applications before the engines are strained. By that I mean 17 dimensional financial apps - pushing the limits of ERP systems to maintain their codeblocks... I've signed up to the Teradata vision of the Active DW, still on the basic hub and spoke model, but with provisions for realtime and the whole 5 level model.
So from my perspective, it's a race to get to EAI and MQ kind of datafeeds into source agnostic, platform-independent, interoperable OLAP tiers. That's something I am happy to see Oracle get into if they can sustain interest in the market. MS will be spending the next two years getting 2000 users to ramp up to .NET and MSSQL2k5, so I don't think there's much hope there. There is a big question as to what improvements are rolled into MSAS2k5.
Meanwhile Hyperion still has the upper hand with integration. The question is whether they can convince 3rd party application vendors to jump on board.
In summary, if Oracle is alive and well and ready to compete platform to platform with Hyperion and Microsoft that's all good, because who knows what IBM is going to do next? Plus it gives me some assurances that Oracle doesn't destroy any more of customer confidence by threatening to swallow Hyperion. Don't bite with your mouth full, Oracle.
So apparently Oracle is facing the same difficulty as Hyperion, which is a broad installation base of their applications with only a few people calling for raw scratch built analytical applications. I basically hear nothing of Oracle OLAP. While there was never much doubt of its scalability and power, there were always questions of whether it integrated well. I've been to admittedly few presentations since 10i rolled out but the entire migration path for Express seemed very confused.
If Oracle OLAP is indeed back in the saddle and well integrated with other vendor front-ends (supporting MDX for example), then that's competition that is welcome. It would consolidate the basic understanding in the industry that OLAP is a platform solution - a comprehensive tier that should interoperate in all environments. MSAS' Windows only implementations muddy the platform message, and the huge ugliness of BO Universes, which are neither totally front-end, nor really ETL, nor as integrated as MSTR further confuse. By accomplishing some clarity in this regard, it pays off to everyone in the industry such that best of breed selections can take place.
Comments