Pretty neat!. So now you can not only create map-based mashups in minutes but also host them.
I could not wait to try this out. So I took the camping map from my last trip (Lake Rooselvelt and Grand Coulee Dam) and decided to crunch the NPS-published high-resolution map that shows all the campgrounds and boat launch information and overlay it on Microsoft Virtual Earth using the tool. Then I created 6 render points and let the mapcruncher algorithm do all the number crunching to figure out how may tiles I need, what tiles I need etc. I locked my images with Microsoft Virtual Earth and saved my mashup (lake.yum). It estimated that I would need 3990 tiles around 340MB of data. Now the best part, I clicked on radio button "Render to S3", it asked me for my AWS credentials and boom! it started uploading all the images/tiles to S3 with just one click of a button. After a minute or so, I got a brand new mashup of Lake Roosevelt and Coulee Dam superimposed on Microsoft Virtual Earth and hosted on Amazon S3.
Obviously, you can see that I am not a expert mapcruncher and the map got skewed to adjust with resolution. Part of the reason was that I only used 6 render points (minimum) instead of creating more for enchanced accuracy. The tool also has features to adjust transparency, which I did not get a chance to play with.
However, the thing to note here is Microsoft Researchers were able to integrate highly-scalable storage infrastructure to their application - mashup in minutes - using the power of Web Services!
Next, I am going to create my neighborhood community mashup where I will overlay all occupant information on Virtual Earth so my neighbors can find other neighbors :)