jeff, thanks for the input. made my trip over to the client site today. Remember the 100mb/s (mega bytes) backup i was told about on the filesystem. Well I had a tablespace with sufficient data put in backup mode. then had netbackup backup the files that belong to that tablespace. What happened? it wasnt 100mb/s it was indeed 50mb/s the same rate as rman. thats 50mega bytes/s.
The interesting thing is, we started with 2 channels, filesperset=1 and maxopenfiles=1 - this gave us 50mb/s - as usual. I then added an additional channel kept the filesperset and maxopenfiles settings the same and that took the rate up to 60mb/s - a 100gb/hr improvement. I then proceeded to open up another channel - 4 channels - and sure enough the rate dropped - seems the threshold is 3 after which io read rates degrade - hence the write rate falls.
This was all done on a reporting db which at the time did not have much load - statspack shows the load on this system is much less compared to production.
We then tried the same backup settings on production and could only throttle at 50mb/s - to me this indicates an io problem? does it not? I cant seem to generate the iostat's myself (not allowed) the system admins tell me iostat results are fine and there are no waits etc. What else could i possibly do to show that the io on the system is being maxed out? and hence rman backups will be slow (well even non rman backups). This is indeed if my hunch about the io is correct .