> I guess I'm confused.  How, exactly, do we measure the maximum aggregate
> operations/sec throughput of the disk?  There should be no upper bound
> on the write throughput unless we reach that number, because writers
> should never be put to sleep.  Is that not the case?
>
> This sort of issue is one reason why I suggested monitoring per-request
> latency as the congestion metric, rather than merely using raw requests/sec
> numbers.
>

We are indeed using and monitoring per process write request as a
metric for measuring congestion. All processes are treated
independently and are put to sleep whevener they are "observed"
issuing write floods.....The uvm_cca code is in genfs_putpages() and
traps all processes (except ioflush and pagedaemon).