@takeshitakenji MySQL is probably the bottleneck, I think.  4 MS/s isn't much disk throughput for a mechanical let alone (I'd think) a SSD.  

What's your innodb_buffer_pool_size?  It's usually the first candidate for optimization.  Ideally you want it to be between 60-80% of available RAM depending on what else you're running on the system.

You might want to consider setting innodb_flush_log_at_trx_commit to 2, this will limit flushes to once per minute which will greatly reduce IO load but has data safety implications.