Philip Howard over at Bloor Research has written an article that discusses the merits of both Exadata and Dataupia when it comes to scaling an Oracle data warehouse.
Registration is required, but it's well worth a read:
My very brief take on Exadata goes like this:
high entry point in terms of database size - you can't buy 'a little'
high entry point in terms of price (see above)
expensive per TB
highly complex i.e. very un-appliance like
only a small sub-set of queries will benefit from 'end-to-end' parallelism - fact:fact joins within the storage layer are not supported
RAC is not an MPP enabler
How this joint Oracle/HP Exadata offering fits with the HP Neoview offering is anyone's guess.