Thanks to all that came to my sessions at VSLive in Orlando on Tuesday, and thanks for the great feedback. Below are the demo's for the two sessions. Updated slides decks will be made available on the attendee website.
Some of you asked which tool I used to demo the perf difference between sync and async (demo 1). Although you could use perf tools that come with VS2005 I used Application Center Test (ACT) that comes with Visual Studio 2003, because it is very easy to setup and shows nice graphics (which was most important for this demo). You can also use the free Web Application Stress Tool. The ACT script I used was (substitute "slow.aspx" with "slowasync.aspx" for the async test), with 50 browser connections:
Randomize()
If Rnd(10) < 5 Then
Test.SendRequest("http://testserver/Demo01/Fast.aspx")
End If
Test.SendRequest("http://testserver/Demo01/Slow.aspx")
Test.SendRequest("http://testserver/Demo01/Fast.aspx")
Note: I've run this test in several setups with different hardware. In my Orlando session I showed numbers from ACT on the host to the web in a VPC. In my San Francisco session ACT was on a Dell D600 with the web running in a VPC on my Dell D800, which is a better way to test (in the Orlando setup ACT is drawing perf from the same CPU). The async perf is much better (even in percentages) with the SF setup. Be aware that there are a lot of variables: CPU capacity, memory, and last but not least... network interface. In another test setup I hardly saw any difference between sync and async. Why? Because the network was the bottleneck :(.