You REALLY need to state EXACTLY what you're going to be doing and why. Especially with hadoop. Moving to hadoop can actually be significantly worse and include complete codebase rewrites if you have no idea what you're doing.
HiveQL is nothing like SQL. Hadoop's method of operation is nothing like RDBMS. Hadoop is not something you can just change a traditional mysql application to.
Additionally, if you are not actually doing big data big data, hadoop is more likely than not a waste of your time because of the complete rewrite and restructuring of everything.
ohhh... ok, sorry i tried to go in the wrong direction.
I'm currently administering a hadop cluster for a customer to store around 2tb of images/videos. I'm doing this for more than 2 years now. Until now we had no un-scheduled downtime and the load on the servers is usually less than 1.
Just to be very clear: you cannot just "replace mysql with hadoop" and solve all your load problems - you now have to rewrite your entire (php?) code for the site itself, as mysql queries no longer work, you need to also maintain new java applications...
I know you might not want to divulge more info about your setup, but switching to something like hadoop really r equires that you do -- it's more than likely if you swap to something like raid10 SSD/FusionIO drives and have better indexes set up on mysql (and use drop-in percona or MariaDB 10), it may fix a lot. Hadoop is not the one size fits all, and if you're using it for like 100GB of data, there's no point in using hadoop in the first place