There's two parts to testing. Testing for quality - which doesn't require the full data set. You can test with 10 products for that. For performance and stress testing, you will want to test with the whole data set.
What specifically are you wondering about with respect to large data?