The frequent argument for Oracle over Sql Server is scale-out ability. With Sql Server's Active-Passive clustering model, the only real way to scale is up with a box with more CPUs, where Oracle's RAC is basically active-active clustering. However, Sql Server's concept of a database is more like that of a super-schema when compared to Oracle's server=database model, giving an additional point of vertical decomposition. In otherwords it would be pretty easy to move a database to a new server should it outgrow it's current shared server. The only problem with that is that if you have a distributed transaction writing to two different databases, how would you resync these databases if you needed to restore them from backup (assuming that it's impossible to have two backups start and end at exactly the same time)?