Well, the second day is drawing to a close. I attended four sessions today, one on data mining, one on data warehousing, one on reporting services, and one on CLR integration. You would think that I would have a lot to talk about, and I do (don’t I always?), but to be honest I am having trouble wrapping my noodle around everything that I absorbed today. I think I am trying to hard to take what I am being given and figure out how to make it best work for my company right now. So, I often find myself thinking about a specific application of a concept being introduced, which causes me to lose focus momentarily, and I have to get caught up from there.
The session on data mining was quite interesting. I have not used any of the data mining capabilities inside SS2K, and was quite impressed with some of what I was shown. My guess is that SS2K5 is even more robust, and we could probably all use a class on analysis services and data mining. The speaker was from Germany and quite knowledgeable on various statistical methods. Being a bit of a math geek myself, I am eager to test out some ideas I have for my fantasy football site. After all, wouldn’t you want to be able to mine data to determine which free agent is most likely to score the most points this weekend? Yeah, I thought so. I am guessing that we have more than one area in our company that could benefit greatly from analysis services, and I hope to learn more in order to best serve those areas and teams.
The data warehouse session was presented by two men from Germany that work for Avanade, which is a joint effort between Microsoft and Accenture. The effort is specific to building data warehouses for companies around the world, and the presentation was delivered as a best practice guide. What they did was to put together a list of pitfalls based upon their experience, as well as where to best focus your energies. What I took from their presentation is that we need a dedicated architecture team to build a data warehouse. Without such a team, then we risk building something that is not only unmanageable, but unusable, as well as impractical. When it comes to a data warehouse, details matter. Such things as accurate metadata pay huge dividends in the end, so it would be prudent to invest the time in getting them right from the start.
The session on reporting services centered around a specific solution that two men from Italy developed for their company. The backbone for their solution is solid, very detailed, but in my opinion the end result is not worth the effort. I am guessing they have just gotten started on this, and will look to continue building their solution until it is quite robust. For example, when asked if a user would be able to view a report they ran last week, the gentlemen said they had no way to handle it right now, but they have some ideas in mind.
The last session, on CLR integration, was very interesting. Supposedly, this is something that I should not fear, though I still have my doubts. The main question I have always asked myself is “Self, when should a developer use the CLR inside SS2K5?”. To date, Self has not been able to answer. Until now, that is.
User-defined functions would most likely perform better in managed code, for example (except for the first time, when it is compiled, of course). So, let’s say that someone wants to create their own aggregate function, named medium(). That would not only work better in managed code, but it is probably not possible in T-SQL. With the CLR, the developer would have access to the rich .NET framework that is available as well, which is a wonderful enhancement to any time someone has struggled to get T-SQL to do the impossible.
Now, while some things should probably be brought inside the database, that does not mean the entire application should be brought inside the database. It would not be advisable to use CLR to created what would essentially be an object oriented database. CLR integration is here to stay, so us DBA folk will have to get used to it, and be able to make recommendations when necessary.
That is all for today. Four more sessions on tap for tomorrow. Until then, adios!
