Do you know the problems, if any, of storing binary data in Riak?
Will it effect the maintainability and gratifaction from the clustering?
What can the performance variations be between using Riak with this as opposed to a distributed file system?
The only issue I'm able to think about is storing binary data bigger than 50MBs that they advise against. The entire reason for Riak is simply that:
One more reason one might pick Riak is for versatility in modeling your computer data. Riak will store data you know it to inside a content-agnostic way — it will not enforce tables, posts, or referential integrity. Which means you can store binary files right alongside more programmer-transparent formats like JSON or XML.
Contributing to @Oscar-Godson's excellent answer, you are prone to experience issues with values much more compact than 50MBs. Bitcask is most effective for values which are up to and including couple of KBs. If you are storing large values, you might want to consider alternative storage backends, for example innostore.
I do not have knowledge about storing binary values, but we have a medium-sized cluster in production (5 nodes, around the order of 100M values, 10's of The best spinner's) and we are seeing frequent errors associated with placing and locating values which are hundreds of KBs in dimensions. Performance within this situation is sporadic - some occasions it really works, others it does not - therefore if you are likely to test, test at scale.
We are also seeing issues with large values when running map-reduce queries - they just break. However that might be less highly relevant to binary values... (as @Matt-Ranney pointed out).
Also see @Stephen-C's answer here
Personally, i haven't observed any issues storing data for example images and documents (both DOC and PDF) into Riak. I do not have performance amounts but might have the ability to gather some must i remember.
Something of note, with Riak you should use Luwak which supplies an api for storing large files. It has been pretty helpful.