I just spent the last week porting over our site from IndexTank to Solr.
I use practically every Amazon computing product already. If it is a Cloud Search Service, then I hope that it is inferior to running you own Solr stack in every shape and form. (Because I love Amazon so much I would proabbly start migrating everything again)./
A few reasons why I've migrated, and I prefix it with saying I've always been a big fan of IndexTank.
1) I was never interested in running my own Search stack. Now forced with the need to run IndexTank myself, this benefit was lost.. Seeing as I'd have to run IndexTank myself
2) Even though it's open source, I doubt that LinkedIn, will be accepting any pull requests.. I imagine that this is nearly the end of the massive development drive that has previously gone on. (Purely my own thoughts, I hope I'm wrong)
3) The faceting with tokenizers is more powerful in Solr. (non existant in IndexTank)
4) The categorization implementation in IndexTank was a little flawed. ie you couldn't create "Grouped" fields, that can contain multiple values. To do this, you had to invert the keys of the values, so that it maps like "Sony=>manufactuer" for example, where as in Solr, you can create manufacturer=>[Sony] and have multi value fields, such as theme=>[thriller,romance] etc.
5) Solr already has a Suggest more like these feature
6) The geo spatial stuff works better in Solr. Especially if you need facets on geo spatial documents to provide counts back.
7) 4.0 (available in trunk) has real time search. So no index rebuilding.
Don't misunderstand me in saying that IndexTank is poor.
IndexTank worked great, and I reall liked it. Perhaps my application just outgrew the features that it could offer and Solr became the obvious choice.
I use practically every Amazon computing product already. If it is a Cloud Search Service, then I hope that it is inferior to running you own Solr stack in every shape and form. (Because I love Amazon so much I would proabbly start migrating everything again)./