Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
Home -> Community -> Usenet -> c.d.o.server -> Imported DB runs SLOWER on faster machine
Hello
I just exported some schemas from an Oracle DB (8.1.7, Solaris) and imported them into a Test DB (Windows NT). Die PC is somewhat faster than the Sun Enterprise 3500 (although it has 4 CPUs , UltraSparc II, and a Raid), as I could confirm doing some tests without the DB.
But some long running applications seem to never end on this PC. A job
running about 4 min on the Sun system seems to run for DAYS not! I was
able to determine that it definitely has something to do with I/O.
Well, this PC has a cheap Promise controller and a Raid 0 system with
4 EIDE disks. As I said, using other applications and benchmarks
proved that the system is really fast (Athon 2000+ MHz, 1.5 GB RAM,
>50 MB/sec sustained transfer rate from the harddisks.
Two possibilities:
1 - the data chunks read by Oracle mismatch the overall stripe size of
the Raid 0 disks. Oracle always reads at least db_blocksize x
db_file_multiblock_read_count. I could not find out the stripe
parameters, because there was no chance to reboot the PC.
2 - Datafiles on the NT system are too large (>20GB). I just forgot that it is not Unix :-) Could anyone comment on maximum Oracle datafile size under Windows NT SP 6? But hey, the DB works somehow.
Anyway, when the application runs, there is ALWAYS contention of type db_sequential_read, and as one updates the number of logical read blocks, it is possible to see how slow this operation works. The CPU is almost idle, contrary to the Raid.
But I still know nothing about the real CAUSE of the problem. Could any one out there give me some hints about how to identify it?
Thanks a lot
Rick Denoire Received on Mon Jul 22 2002 - 18:12:23 CDT