Operating System - HP-UX
1751725 Members
6086 Online
108781 Solutions
New Discussion юеВ

Re: Oracle db perf on AUTORAID

 
Ponniah
New Member

Oracle db perf on AUTORAID

We have the following setup:
Server1-HP01 -
8cpu/16GB RAM - Datawarehouse database(PROD)

Server2-HP02 -
4cpu/4GB RAM - Dev (5 OLTP + 3 DSS) databases

Server3-HP03 -
8cpu/8GB RAM - OLTP database (PROD)
Server 1&3 are in cluster acting as failover for each other.
All the 3 server share 2 VA7100(RAID0+1) currently.
The production data is also mirrored between the VA's thru LVM mirroring.
We were planning to go for AUTORAID
Can you give us any suggesstions / implication for using AUTORAID for this?
TIA
Shiv
environment.

Shiv
5 REPLIES 5
Jeanine Kone
Trusted Contributor

Re: Oracle db perf on AUTORAID

I recently switched my va7100 from raid 0/1 to autoraid. I have to say, I did notice a slight decrease in performance. Not enough for me to really be concerned about (for example, mt full export increased from 3.5 hours to almost 4 hours).
Ponniah
New Member

Re: Oracle db perf on AUTORAID

Thanks. Any impact on imports / batch process / ETL loads?
As we are in development stages yet there are extensive of exp/imp/refresh/dataloads on multiple databases. Do you think the performance will be affected with autoraid?
TIA
Shiv
Shiv
Hai Nguyen_1
Honored Contributor

Re: Oracle db perf on AUTORAID

VA7100 is actually an AUTORAID array. If data are currently referenced then they will be in RAID 0+1. Otherwise, they will be in RAID 5 double parity.

Hai
Ponniah
New Member

Re: Oracle db perf on AUTORAID

Hi,
As of now we have put on hold the idea of AutoRAID due to performance consideration.
Thanks for your responses.
Ponniah
Shiv
Jeanine Kone
Trusted Contributor

Re: Oracle db perf on AUTORAID

Again, just based on my recent experience - my import time did increase as well. Out import into the test database which normally takes about 36 to 40 hours, increased by about 4 hours. I have not tested any of the timing for the day to day OLTP processing. I have not heard any complaints from the customers yet (it has been almost a week since the switch on production).

I do think that the performance impact will depend upon how mcuh data you have on the array versus free space. If you have a lot of free space (some of which in unallocated I believe), then the most active data should remain in RAID 0/1 with better performance. If you have very little free space (like us) then you will not be able to keep all of the very active data in RADI 0/1 - and some of it will have to remain RAID5 with the slight performance hit.