Take any handy terrabyte size database, preferably with one table that comprises a significant part of that terrabyte. Telco data or extremely high volume transactional data are the only real good candidates I know of for this kind of demonstration.

Create a prep table of a few million rows with ordinary (non-sequential) GUIDs.

Start the timer

Append the prep table to the main table.

Stop the timer

Repeat the process with sequential GUID values in both the main and the prep tables.

I don't have spare hardware to create and test a demo script for you. Please let me know how the sequential GUIDs perform versus the non-sequential GUIDs in your test. I know of three companies (one was Microsoft) that have run this test, and all three of them found the non-sequential GUID test to run significantly faster for all cases except the initial table load (the first 100 Gb of data).

-PatP