RE: To ODA or Not?

From: TJ Kiernan <tkiernan_at_pti-nps.com>
Date: Fri, 27 Mar 2015 21:26:16 +0000
Message-ID: <196DB2D4BDE5804EAF3158CCC1C698BC8BEAA167_at_lopez.pti-nps.com>



The disk groups all use the same underlying storage pool, so long as you’re using the hardware in appliance mode (I *THINK*). You can reconfigure the hardware and assign physical disks to storage pools, but the default is one big mass of disk that ASM diskgroups, volumes, acfs mounts, etc spread out across. This is the impression I had from the Oracle sales guys I talked about, which made me think twice about consolidating OLTP & DW instances onto a single ODA.

***usually I wouldn’t be posting with such uncertainty, but given the lack of responses on the last thread, I figured it’s worth mentioning for followup with someone who knows for sure.

HTH,
T. J.

From: Jeff Chirco [mailto:backseatdba_at_gmail.com] Sent: Friday, March 27, 2015 4:21 PM
To: TJ Kiernan
Cc: oracle-l_at_freelists.org; MARK BRINSMEAD; Seth Miller; jack.applewhite_at_austinisd.org Subject: Re: To ODA or Not?

I was under the impression that you could carve out separate sections of the disk groups to allocate to different areas. Correct me if I am wrong on this. However our current SAN is set up like with with just one aggregate. So yeah sometimes when I run an massive query in our DW, our other database get affected. Oh and I forgot to mention that when I ran dbms_calibrate_io I got back 3400mbps of IO with about 11ms of latency. Not sure what you get on the ODA storage.

On Fri, Mar 27, 2015 at 2:05 PM, TJ Kiernan <tkiernan_at_pti-nps.com<mailto:tkiernan_at_pti-nps.com>> wrote: I’ve been glancing at the ODA since version 1, which was our last hardware refresh. We ended up rolling our own, but we’ll take a serious look again when we outgrow the existing kit. (3 years ago, you could buy a box with only 4 cores. I’m not sure if that’s even possible still.)

The one concern I would add to your current list is I/O. If I recall correctly, the storage is one big mass of disk (plus the SSDs for redo). If someone kills the I/O channels in your dev environment, can it adversely affect any other (such as production) databases running on the ODA?

Maybe that’s not a concern with your systems, but it’ll mean one ODA for our busy database that makes the money plus one at the DR site.

Thanks,
T. J.

From: oracle-l-bounce_at_freelists.org<mailto:oracle-l-bounce_at_freelists.org> [mailto:oracle-l-bounce_at_freelists.org<mailto:oracle-l-bounce_at_freelists.org>] On Behalf Of Jeff Chirco Sent: Friday, March 27, 2015 3:39 PM
To: oracle-l_at_freelists.org<mailto:oracle-l_at_freelists.org>; MARK BRINSMEAD; Seth Miller; jack.applewhite_at_austinisd.org<mailto:jack.applewhite_at_austinisd.org> Subject: To ODA or Not?

I had another thread started about the ODA but I wanted to separate it out. I am currently up for a hardware refresh and tossing around the idea of Oracle Database Appliance. But in my environment I am not sure if it just that easy. To start, everything currently runs on Windows 2008 R2 64bit. So that is our first big change, going to Linux. We have 4 servers:

  • Production 1: Runs 5 EE 11.2.0.4 database's but really 4 of those are mostly idle with one acting as our main system. This a 4 CPU box so with a .5 multiplier we are licensed for 2 CPU. And currently we are not even utilizing half of the CPU and about half of the memory which is 96gb.
  • Test 1: Same server and license as above it just runs all test and dev databases
  • Production 2: Runs 3 SE1 11.2.0.4 database for third party applications but the communication to Prod 1 for some data. CPU is also lightly used on here.
  • Test 2: 1 SE 11.2.0.4 database on here but this server is not that important.
  • 5th server acts as our Data Guard server for just 1 of the database running on Prod 1. This is a single CPU server. Meant as a last resort.

Current Storage: NetApp SAN, 27tb. The NetApp SAN is nice with ability to quickly take a database snapshot/backup (1-2minutes) and then make a database clone from that in less than 5 minutes for our 400gb database. Plus it has the ability to Snap Mirror all the snapshots to our off site location.

If we were to get an ODA since we don't run or have a current need for RAC,

  1. One solution would be to have one of the servers for Production and other for Test/Dev. I would combine all of our databases into this one machine. Then they would all be EE and probably go to 12c and plug able if I can get the budget approval for it. But there is another added expense.
  2. The solution I am thinking makes most sense is turning VM on kind of mimicking our current environment so I can keep our third party systems separate and on a separate database version if need be. But still all on Linux. Then I can create another VM for Enterprise Manager which I currently have running on Windows.

But what do I do about our existing snapshot technology with NetApp and the mirroring. I believe there is not data mirroring technology with the ODA storage. Then are we relying on Data Guard everything or moving backups some other way? I heard the ODA has the snap clone technology so we could still create dev clones quickly right? Is this an extra cost?

Speaking of Data Guard, since we probably wont be buying two ODA's, we will have to stand up our own stand alone Oracle Linux server and support that and buy support for it.

Speaking of buying only 1. I dont't think I can get approval for that unless they give it to us dirt cheap. This means we have no way to test out ODA patches and updates. This sounds scary to me.

Ok so that was a long email. Thank you in advanced for anybody that got through it and chimes in and I'm aware that this probably requires a deeper discussion.

Jeff

--
http://www.freelists.org/webpage/oracle-l
Received on Fri Mar 27 2015 - 22:26:16 CET

Original text of this message