Sounds reasonable. How big are your writes? Also are you seeing a bottleneck? If so, what are the details? If everything is running fine with 200k writes/sec (compactions aren't backing up, not a lot of disk IO) then that's good. However you will also need to compare what you can achieve when you actually want to read the data, as reads are much more expensive than writes and it may greatly impact the number of w/s you can achieve.
You can tune Cassandra a lot to squeeze out as much performance as possible, but really depends what you're doing with it. Worth getting a benchmark of the load you expect to be on the cluster (both reads and writes) before going down a tuning path.
