UNIX gurus: DB performance on Solaris

[DBAtrix]

Those of you who moonlight as sys admins:

I set up a new develpmt database on a different server. I set up the database based similarly as possible on another database: init.ora/config.ora params are the same, temp tablespace size & extents are the same, RBSs are built the same. The data segments are the same--both are imported from the same production export.

Server differences:
Server A
------------------------------
Machine type: sun4u Sun Ultra 2 UPA/SBus (2 X UltraSPARC-II 296MHz)
CPU: 2 x UltraSPARC-II at 296 MHz
Memory: RAM : 512 MB
Host adapter: fas0 fas Qlogic FAS366 SCSI 2 Fast/Wide
2 hard drives

Server B
------------------------------------------------------
Machine type: sun4u 4-slot Sun Enterprise 3000
CPU: 2 x UltraSPARC-II at 248 MHz
Memory: RAM : 3072 MB
Host adapter: fas0 fas Qlogic FAS366 SCSI 2 Fast/Wide
6 hard drives

Databases on both servers run EE 8.1.5.1.0

The import takes 4 hrs on Server A. The import takes 7 hours on Server B (again, imported from the same source). I used the same import command line (indeed, i copied the parfile). I did not expect such a difference in performance. Pertinent details from the import: 1.) using < buffer=200000 commit=y > 2.) the bulk of the import is consumed by one table, with a long column.

Please note I am *not* looking for performance recommendations for import per se (I have experience in that realm); I am looking for UNIX expertise to help me account for the import time difference (4 hours vs. 7 hours) in what I expected to be similar results. My experience on the sysadmin side is quite limited. I don't know where to look for what would yield this difference, with the exception of /etc/system file, but i'm not sure how that would affect it.

Thanks for your insight, D.

[Edited by DBAtrix on 11-17-2000 at 05:09 PM]

[marist89]

First thing I would do is check to make sure the /etc/system files match up.

Next, I would look at the IO. Is system B's datafiles spread out similarly to system A? Is System B Software RAID 5 where System A is straight disk?

Next, I would use some OS utilities to see what's getting swamped:
iostat -xdM for I/O
sar for CPU, Swap, & Paging activity

Last, I would look to make sure my network connectivity is being fully utilized (if importing over net). If you think this might be a problem, imp from the local machine.

[Chandra]

You have you specify the O/S details.
ie. what file system. how did you distribute your datafiles.
and check the I/O.
you can find this information by using the "top" utility.


Good Luck

[DBAtrix]

I have more diskson Server B (slower import) so i have defacto better i/o distribution. (straight disks on both servers A & B, no RAID)

Thanks for the suggestions.

D.

[pwoneill]

Is it more disks on one controller or multiple controllers? Also are the files configured the same to minimize contention?