I am running a simple 3 node cluster.
Each node is relatively powerful. (16 core, lots of ram, good disk etc).
All machines are identical.
Token ranges are all equal (generated with the python script on datastax website)
Although under heavy load I can see the "unix load" be rather uneven between the machines. (ie one machine may be 1/5th of the load of the others.... double digits vs single digits).
Where would you recommend me to start tracking this down?