I can't see what was wrong with the second graph. Regular bursts of writes for flushing leaves the disks more idle for low-latency reads. Optimising throughpu

I can't see what was wrong with the second graph. Regular bursts of writes for flushing leaves the disks more idle for low-latency reads. Optimising throughput doesn't mean you have to have a smooth, flat write graph. (For comparison, see ZFS)