You should check out Lightcrest, they provide high-availabilty hosting infrastructure - Amazon EC2 and Rackspace are incredibly expensive when you take into consideration steady-state load.
We were actually considering both an initial crawling mechanism so you can "just go", and a very simple REST API for adding, updating, and deleting existing documents in the index. So you would be able to have both a 0-60 approach for getting your already-public-facing content into a sub-second search index, and also the ability to granularly submit document batches.
As far as security is concerned, one assumption is that if your content is already crawlable, it's not private. That said, it would be trivial to implement a mechanism for keeping private data private - and even public data from being indexed (i.e. with a meta tag.)