Unsolved
This post is more than 5 years old
225 Posts
0
1640
Ora Layout question, about EFD and F_cache.
Greeting,
I like to know which is better for oracle performance with specific # of EFD in a OLTP environment
1. Manually put Oracle Redo log and Undo tablespace files on EFD.
2. Put whole DB on a FC pool with F_cache enabling.
I think F_cache would be better since it would actively select hot data, although Oracle Redo log and Undo tablespace files are most hot part in a OLTP DB.
dba_hba
63 Posts
0
September 6th, 2011 04:00
Eddy
I have been contributing to an, as yet, unpublished whitepaper testing FASTCache. Best if I include a couple of quotes from it
1. "Oracle archive files and redo log files have a predictable workload composed mainly of sequential writes. These can be efficiently handled by the array’s write cache and assigned HDDs. Enabling FAST Cache on these files is neither beneficial nor cost effective."
Similarly there is little or no benefit assigning Flash to your redo when it is handled by an array’s write cache .
2. "In line with EMC recommendations, FAST Cache was enabled for the Oracle data files only."
Also you can look at the following whitepaper on Powerlink
Oracle database on EMC VNX Best Practices
http://powerlink.emc.com/km/live1/en_US/Offering_Technical/White_Paper/h8242-deploying-oracle-vnx-wp.pdf
hope this helps answer your question
reseach
225 Posts
0
September 7th, 2011 03:00
Allanr, thanks for you helpful information.
I have a bit further question on this thread.
As my understanding on Redo, every modification on data files would generate at least 2 updates on Redo , although DB write process would append redo file in seq write, every single SSD provide closely 300MB/s through put, about 6-7x faster than SAS spindle.
Any comments?
Thanks,
Eddy
dba_hba
63 Posts
0
September 7th, 2011 07:00
As per my previous reply,
"Similarly there is little or no benefit assigning Flash to your redo when it is handled by an array’s write cache"
Allan
reseach
225 Posts
0
September 8th, 2011 03:00
I do not disagree with this part. I just think it is a bit idealized. VNX SP cache is running within “high water mark” and “low water mark”, if amount of Redo file update is not huge, SP cache is good, otherwise, it would drive it into “forced flush”, that could affect host performance.
EFD performance is more sustained, considering cost, SAS spindles w/ SP write cache is relatively good choice.
Thoughts?
Eddy
dba_hba
63 Posts
0
September 16th, 2011 04:00
Eddy
here is a great post by EMC's Kevin Closson which discusses redo logging in depth and should address your question regarding redo and flash
http://kevinclosson.wordpress.com/2007/07/21/manly-men-only-use-solid-state-disk-for-redo-logging-lgwr-io-is-simple-but-not-lgwr-processing/
reseach
225 Posts
0
April 24th, 2012 03:00
today I looked @ ODA configuration, it is with 292GB Flash for redo log. comments?
dba_hba
63 Posts
0
April 24th, 2012 03:00
Yes
ODA has no array cache
https://community.emc.com/message/563946#563946
This url is for Exadata but I hope you will pick up the general idea
http://jarneil.wordpress.com/2012/03/20/exadata-smart-flash-logging/
reseach
225 Posts
0
April 25th, 2012 18:00
Ok, got it, that makes sense.
I was wrong, I thought ODA storage SN was a server, but no, it is a shared SAS JOB.
I find a document describe its HW components well, see the link, www.artimbilisim.com/bulten/ODA-VAD-VAR-Tech.pps
Eddy