Email Subscription Notifications Suspended Temporarily
We are in the process of making navigation in the Servers and Operating Systems forums simpler and more direct. While doing this, we have to temporarily suspend email notifications for subscriptions. If you are subscribed to one or more discussion boards or blogs in the community, please check them daily to see new content. Notifications will be turned back on in a few days. We apologize for any inconvenience this may cause. Thanks, Warren_Admin
Disk Enclosures
cancel
Showing results for 
Search instead for 
Did you mean: 

EVA4000 question

SOLVED
Go to solution
Charles McCary
Valued Contributor

EVA4000 question

Group,

hi - looking at purchasing an EVA4000. Our array use is primarily oracle database and significantly heavier on the physical reads than the physical writes. I've been reading up on the best practices for this array thinking about performance and availability... and given that, here's what I was thinking for configuration:

1) Two disk groups, one for data and apps, one for logs.

2) Use Vraid1 for performance and availability for all LUNS.

3) I'd like to use multiple of 8's in the disk groups, but I'm limited to the amount of disks that I can purchase. I may be able to swing this though and if I can I will.


Any thoughts on the above or other suggestions?


I'm hoping to take advantage of 4gb from host to disk and improve performance. Right now we're on an EVA3000 with 2gb fiber cards and we're using Vraid5 and we really didn't put too much thought into the disk groups....


12 REPLIES
Court Campbell
Honored Contributor

Re: EVA4000 question

This may help you make your decision.

http://www.oracle.com/technology/deploy/performance/pdf/EVA_ORACLE_paper.pdf

"The difference between me and you? I will read the man page." and "Respect the hat." and "You could just do a search on ITRC, you don't need to start a thread on a topic that's been answered 100 times already." Oh, and "What. no points???"
Charles McCary
Valued Contributor

Re: EVA4000 question

Interesting...they see no real advantage of Vraid1 over Vraid 5 in throughput. HP's best practices indicates differently....good doc.

thanks,

c
TwoProc
Honored Contributor

Re: EVA4000 question

I've got a "white paper" of sorts that I've supplied the HP/Oracle team with over the years that covers this very topic. It's not as polished as a "white paper" - it's more grey... I've asked on of the HP team member for a copy which I should be getting soon (my personal HD crashed a while back w/ no backup, and that's really sad for an admin to use as an excuse - 'cause it's not one). I'll put up as an attachment to you as soon as it shows up.
We are the people our parents warned us about --Jimmy Buffett
tkc
Esteemed Contributor

Re: EVA4000 question

you may refer to white papers in the following link :

http://h18006.www1.hp.com/storage/arraywhitepapers.html

there is one document on best practices for EVA4000/6000/8000
Rob Leadbeater
Honored Contributor

Re: EVA4000 question

Hi Charles,

If your database is significantly more reads than writes, then your logs probably aren't going to being used that much. Therefore you *may* be better going for one disk group rather than two.

That will also make the bean counters happier as you'll get more disk space for your money. With two disk groups, you'll be losing the capacity of at least 4 disks...

I would probably stick with Vraid1 if you can afford it. You don't say how many disk shelves you're looking at getting, however there is greater risk of losing Vraid5 over Vraid1 in the event of multiple simultaneous disk failures.

Hope this helps,

Regards,

Rob
Mark...
Honored Contributor
Solution

Re: EVA4000 question

Hi,

If you are looking for maximum availability then you would be best going for 8 shelves with multiples of 8 disks in a disk group[s] spread vertically over the disk shelves. This is known as a Robust Availability Config [RAC] and is what HP recommend.

A 4k & 6k are the same controller but it will probably cost a bit more. Shelves are not too expensive but you will also need to purchase 2 backend loop switches for a 6K 2C8D and these will push the price up abit!

However this will also give you a better / less disruptive upgrade path should you wish to add disks in at a later date - remember multiples of 8 installed virtically across all shelves and all the disks [multiples of 8] added to the same disk group to maintain your RAC.

You pays your money...

Mark...
if you have nothing useful to say, say nothing...
Charles McCary
Valued Contributor

Re: EVA4000 question

Thanks for all of the input. One other question...because of how this array works am I buying myself anything by striping at the LVM level as well?
Court Campbell
Honored Contributor

Re: EVA4000 question

the only advantage I can think of is that it may help with hot spots in the disk group.
"The difference between me and you? I will read the man page." and "Respect the hat." and "You could just do a search on ITRC, you don't need to start a thread on a topic that's been answered 100 times already." Oh, and "What. no points???"
Uwe Zessin
Honored Contributor

Re: EVA4000 question

The EVA is supposed to be able to remove them by relocating data to less busy disks. How small that granularity is, I don't know.

The only argument I've learned is if your SCSI queue is too small and you don't want / can't increase it. By using multiple PVs with LVM striping, you have more parallel IOs.
.
TwoProc
Honored Contributor

Re: EVA4000 question

Ok, readable "grey" paper (not quite a white paper)... attached. All caveats apply, some disagreements in theory with long term HP forumers with leafy hats and excellent reputation are probably contained within - so use what you may of the recommendations, feel free to disregard the rest!
We are the people our parents warned us about --Jimmy Buffett
Charles McCary
Valued Contributor

Re: EVA4000 question

THANKS!
Charles McCary
Valued Contributor

Re: EVA4000 question

n/a