Only if the data within that video begins and ends on the block boundry.

So say you take a 100 MB video, and break it up into 1MB blocks. You now have 100 blocks of 1MB. Copy that file. It is 100% identical, so you will get great dedupe. Now, take your copied file, add 10 extra bytes of data to the beginning. This shifts everything 'down' by 10 bytes. Now, all your blocks of the copied file are 'offset' by 10 bytes from the original, and not 1 of the blocks are the same as the original. By adding 10 bytes, you've killed the dedupe. Further, because its compressed video, none of the blocks WITHIN the file are the same (that was the point of the compression), so you gain nothing there either.

Now, itsnot quite as bad as the example I just gave, but this is a fundamental problem with all fixed-block deduplication efforts (like NetApps). Other implementations dont have this problem because they use sliding windows for their block sizes (DataDomain, shortly ZFS), etc.






--Matt
VCP, vExpert, Unix Geek

Also to go along with what MGower says it's almost impossible to de-dupe a picture. If you copy a picture, and modify it, ALL the blocks in that picture have changed, because how can you tell the difference in the way a picture looks unless you visually look at it. So pictures are the most difficult to identify duplication, so compression is one, but pictures are very hard to compare.