How long does the IFS checkpoint processing take vs the save port in IFS SWA? Was: Disks are getting hammered ?SMPO0001?

What I'm reading is that save while active, for the IFS, skips any file that is currently locked. Unlike qsys.lib objects. The use of SWA for IFS is to get a snapshot of all those unlocked objects before the save starts. The way I think the design concept is set up is that: 1 - Stop all use of the IFS area you're trying to save. 2 - Start the SWA. 3 - When you get the checkpoint message then start allowing people to use the objects again. The save then saves from that checkpoint. So, in theory I could quiesce Domino and that netserver share, start the SWA, monitor for the check point message, when that occurs start up Domino and the netserver share. Where reality steps in is if it takes almost as long to get to the checkpoint (really fast save media, etc) as it does to do the full save then doing the checkpoint is just doubling your save time while leaving you the same amount of time of outage (waiting on the checkpoint instead of just doing the save). What is unknown is just how long it takes to get to the checkpoint time.
If you're NOT monitoring for the checkpoint message and then starting up your operations then you've just wasted time.

Doing two different IFS saves during the run.

First IFS save
23:55:09 Starting save of list *LINK to devices *MEDDFN.
0:09:19 Save-while-active checkpoint processing complete.
...
0:09:20 115680 objects saved.
0:11:37 Save of list *LINK complete.
I find it hard to believe that the checkpoint processing took 14 minutes and 10 seconds and to save 115,680 objects took 1 second. I think IBM is holding on to messages or some such thing.

Second IFS save
0:11:38 Starting save of list FILESHARE to devices *MEDDFN.
0:11:38 Message queue FILESHARE not found.
0:47:15 SAVACT(*SYNC) used with media definition.
0:47:16 Save-while-active checkpoint processing complete.
...
0:47:17 8 objects saved.
0:47:18 Save of list FILESHARE complete.
Again, I find it hard to believe that the checkpoint processing took about 36 minutes and the actual save only took 1 second. Again I think IBM is holding on to messages or some such thing.

Can someone explain this?

Rob Berendt