I was planning on writing another article about partitioned views to provide a guide for getting the most performance from them.
My article on the speed of partitioned views was intended to just state the facts as tested. For example I don't see any point padding out the article with definitions of member tables as these are almost identical to the original table. (Only names of tables, contstraints are different. Indexes have the same names. Check constraints added for partitioning. Also everything is in BOL.)
I gave an example of the type of script used for testing and where I modified for additional tests. I can supply all scripts if you would like but the outcomes are as stated in my article.
The outcomes illustrate the differences in execution times between original table and equivalent partitioned view where number of records in each are 975000000.
The bottom line with local partitioned views is that you need to specify the range of the partitioning column in order to get speed performance improvements.
I hope this will assist people who may be interested in LPV as a solution to certain problems with very large tables.