the "perfect" max load should be the number of cpus you have,
       so a quad-core server can sustain a load of 4 without any
       problem... after that number, the higher the load, the higher
       will be the performance lost

       a high load and still having cpu idle is a sign of IO load.
       high system cpu usage is a sign of the kernel spending too
       much time managing the IO

       for IO load can be 99/100%, but only as long the IO wait
       doesnt increase alot... this values depends of the HD type,
       raid layout, number of disks, bus, etc

       in modern HDs (sata, scsi, sas) IO waits between 5-20 are
       normal... 100 is high load...if you see times above 150,
       that HD is probably getting too much load and is slowing
       down a lot the rest of the system (but again, this values
       depend of your OS/config/FS/HD layout)

       dont forget that writes usually put more load on a HD than
       reads, the HDs cache helps a lot, but usually write cache is
       not enabled

       use iostat (like iostat -kx 30 ) to monitor the HD loads
       and see how the io load and wait is going... sporadic
       high loads/times are normal, specially in small check times,
       but sustained high loads/times are a sign of problems

       if you want to try to get a little more from your HDs and
       the data isnt critical in case of a powerloss (usually
       backups arent, next backup would "fix" the bad data), you
       can enable the write cache.

       be aware that at least xfs doesnt like to lose data with
       powerloss+write cache, its format assumes that data hits
       the HD when it tells and can miss behave if the cache
       just decided to only really write half of the data before
       the powerloss. that is why is highly recommended to use xfs
       with a UPS and do a controlled shutdown in case of powerloss