Your "not loves" don't ring true for me though- perhaps you'd like to elaborate there?
I've been running production apps on Heroku for more than 2 years now, and there has only been one occasion where their uptime impacted the apps' availability. Even during Hurricane Sandy, my apps were up 24/7, so it's hard to understand your comment.
Yes, they do have non-critical issues at times (their API goes through brief, but not infrequent periods of instability)... but this only impacts your ability to deploy/manage an app, not the app's actual uptime.
What is the most performant (but useable) filesystem for petabyte-scale parallel processing, that is running in real Production environments today?I assume that there are newer options than Hadoop that compare well? What I'd really like is a properly-tested Haskell solution, but I imagine that's a few years off yet...
If that sort of thing floats your boat, you should check out Yesod as well (Haskell).