Biased Benchmarks or Cool Code?
The Microsoft keynote...always a good place for grand exaggerations. There were two topics of note here.
First, they talked about .NET usage in SQL Server 2005. They wrote a "stored procedure" into a Delphi.NET class, which rifled through some rows, parsing for a particular substring in a column. They compiled the assembly (dccil has a --clrversion switch where you can target any CLR you want) and uploaded it to the server. They also had the "equivalent" implentation in T-SQL. When run (including a cache clear to remove any accidental optimization) the .NET implementation took ~200ms and the T-SQL implementation took about ~20,000ms. I don't think I trust those numbers, and hopefully Bob Z and I can do some experimentation.
That took long enough that they went through a hurried demo of Team System with Caliber. It looks like Caliber requirements are first class citizens in VSTS. Too bad we won't be using them.
First, they talked about .NET usage in SQL Server 2005. They wrote a "stored procedure" into a Delphi.NET class, which rifled through some rows, parsing for a particular substring in a column. They compiled the assembly (dccil has a --clrversion switch where you can target any CLR you want) and uploaded it to the server. They also had the "equivalent" implentation in T-SQL. When run (including a cache clear to remove any accidental optimization) the .NET implementation took ~200ms and the T-SQL implementation took about ~20,000ms. I don't think I trust those numbers, and hopefully Bob Z and I can do some experimentation.
That took long enough that they went through a hurried demo of Team System with Caliber. It looks like Caliber requirements are first class citizens in VSTS. Too bad we won't be using them.
0 Comments:
Post a Comment
<< Home