Libor is right. I can't believe I didn't see that...
(Of course that means that you should have a good backup first. Or be willing to throw away the target.)
The Furgal Test:
Procopy session begin for root on batch. (451)
Database copied from /apps/dlc117/sports. (1365)
Procopy session end. (334)
OpenEdge Release 11.7.2 as of Tue Oct 24 18:20:59 EDT 2017
Before-image cluster size set to 16384 kb. (1620)
-rw-r--r-- 1 root root 8192 Mar 8 11:20 sports.b1
9 seconds = 10MB/sec -- which is barely acceptable, anything longer and your disk susbsystem is junk.
5 seconds = 20MB/sec -- "ok"
3 seconds = 30MB/sec -- "good"
1 second = 100MB/sec -- which is "excellent!"
So it is also inconsistent. Or getting worse under load as the day goes on.
> So it is also inconsistent.
First run - 18.387 sec
Run as syncio.sh - 18.161s.
But 5 MB/sec is very slow.
I ran the dump and load using the Dictionary Dump, Bulk load and Index Re-Build and here are my results on the same DB and same server and same Disk Sub System.
Dump - 15:36 - 17:47 - 2 Hours 11 Minutes.
Load - 17:58 - 19:26 - 1 hour 28 Minutes
Index Rebuild - 19:45 -20:31 - 46 Minutes
Can somebody explain the difference in time for the Binary Load that took 11 Hours 49 Minutes compare to the Bulk Load that took 1 Hour 28 Minutes.?
Is your bulk load test using a new DB or did you procopy empty into the pre-grown shell of the first test? There may be a cost to growing the data files.
Did you try the binary load with -r ? -r gives you the benefits of -i (FS cached writes) without some of the drawbacks.
@Mike - I will need to re-run the Binary Dump and Load again. Will try this now. Will run the load with -i and -r. I am using the same DB with the same data so this gives me consistency.
@Paul - I created new DB's and the extents grew as I ran the load. But this could not have an impact on Binary Load of 11 Hours 49 Mins compared to Bulk Load of 1 Hour 28 Mins.
My opinion is it seems to be a Progress 11.7.2 issue.!!
mfurgal wrote that bulk load runs with -i so that seems to be the most likely explanation for the discrepancy.
If you have plenty of time, I think that running a dictionary load with all indexes deactivated should simulate a bulk load without the -i.
In summary, these are the various combinations:
1. Binary load w/out -i / -r: 12 hours
2. Binary load w/ -r: TO DO
3. Bulk load (default w/ -i): 2h
4. Bulk load w/out -i (simulated by dict load with inactive indexes): TO DO
I added the -i -r option on the binary load and i have the following results.
Binary Dump - 25 Minutes
Binary Load - 50 minutes - Time down from 11 Hours 49 Minutes to 50 Minutes.
Index Rebuild - 52 minutes
That makes a whole lot more sense!
Thank you all for all your input. It has been a day of learning.