TechEd Memoirs

TechEd 2006 US is over. Given the fact that VS.NET 2005, .NET 2.0, and SQL Server 2005 have matured, there weren’t many earth-shuttering announcements and cool demos. Anyway, here are some of the most interesting BI highlights written in the plane on my way home.


Brian Welcker told us that RS SP2 (CTP scheduled for November, final release alongside Office 2007 launch) will bring a much tighter level of integration between RS 2005 and SharePoint 2007.  This is great given that SharePoint is the Microsoft premium portal solution. Specifically, it would be possible to configure RS in a SharePoint configuration (SSP) mode using the RS configuration utility. In SSP mode, the entire RS catalog is migrated to the SharePoint database. Instead of the Report Manager, the SharePoint Administration portal will be used for report management. For example, the Report Server will authenticate the users using the SSP security settings. The SSP Report Viewer (a wrapper around the ASP.NET Report Viewer) will make the report viewing experience more SharePoint-like. For example, the report parameters will be presented in a dropdown list to minimize the horizontal space.


Jamie MacLennan from the SSAS team did a great presentation about driving report payloads when UDM is used as a data source. What impressed me most was that Jamie gave us an honest account about some of the problematic integration areas instead of a let-me-show-you-how-cool-everything-we-did-is type of a presentation. For example, he pointed out that SSAS hierarchies don’t have Report Builder equivalents. Not only parent-child dimensions are not supported in the Report Builder, but they will give wrong results. He recommended avoiding parent-child hierarchies in the Report Builder whenever possible or replacing them with flattened regular hierarchies in UDM. Jamie shared great tips about associating a calculated member with a measure group so it appears in the Report Builder and disabling the automatic Report Builder drillthrough reports when UDM is used a data source for performance reasons. The biggest surprise for me came when he stated that Microsoft now recommends splitting a large UDM in smaller models that leverage the linked dimension and measure group features. As performance tests by the SSAS team have demonostrated, this approach may result in better memory utilization and performance, as well as easier management.


T.K. Anand and Dan Battagin did a great demo of SSAS-Excel 97 integration. No big surprises here if you follow the David Gainer’s blog. T.K. Anand and Christian Pesculescu took us behind the scenes of the UDM storage model and shared very interesting details about how dimension and measure are stored. Later, they did another excellent presentation that focused on maximizing the UDM performance. They emphasized the importance of designing proper attribute relationships and avoiding duplicate attribute relationships. It was mentioned that there will be a SSAS performance whitepaper in near future.


I was also given a personal sneak preview of the forthcoming PerformancePoint server. As I noted in a previous blog, PerformancePoint will unite three products — Scorecard Manager, ProClarity, and Biz#. I asked the presenter to focus on the Biz# portion only. It realized that the client interface will be implemented in Excel 2007. In fact, it seems like the Microsoft Office team expects people to build BI-centric applications inside the Office products after retiring OWC. The Performance Point client will support comprehensive models for planning and forecasting. Behind the scenes, these functions will be carried out by the server layer of PerformancePoint which creates SSAS cubes to save the model data.


Finally, I think my presentation (Building Reporting Solutions with SSRS and SSAS) went really well. Hilton Logan (the hotel I stayed) surprised us with a false fire alarm in the wee hours of the morning (around 3:30 am) and decided to evacuate everybody outside. So, I went on stage quite sleep-deprived since I couldn’t get more sleep after the accident (oh, yeah, we had an MVP party the previous night). The presentation took place in one of these vast conference rooms that make me proud that I am an American. It could easily take 500 people and it had four large screens. Despite the early hour (8 am), about 200 people showed up to boot with a healthy dose of Microsoft BI. I got a few interesting questions during the presentation and after it in the Business Intelligence discussion area.


As always, TechEd was a great show. Some 12,000 people attended the conference and it was sold out well in advance. I liked Boston too as I enjoy all cities where one can walk instead of sitting paralyzed in a traffic jam. Unfortunately, Atlanta is not one of them 🙁