Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Mailing Lists -> Oracle-L -> Fwd: Question: Oracle Backup to NAS Device SATA vs.. SCSI vs. $$$

Fwd: Question: Oracle Backup to NAS Device SATA vs.. SCSI vs. $$$

From: jungwolf <spatenau_at_gmail.com>
Date: Tue, 15 Mar 2005 17:27:36 -0600
Message-ID: <10644b9e05031515275b2758f7@mail.gmail.com>


Jeff,

Private response was bounced so I'll post this to the list. I'm curious on how "on the cheap" anyone can do Oracle. For example, I've heard about RAC on two laptops and a shared firewire drive. Have any brave (foolhardy?) DBAs put this class of solution into production?

On Tue, 15 Mar 2005 11:23:40 -0600, Henslee, Jeff <jhenslee_at_wausauwindow.com> wrote:
> Environment: Solaris 8, Oracle 8.1.7.4 (250+GB), Recovery Manager, =
> Legato Networker, AIT-3 Tape
>
> Question:
> We're looking to speed up our backup window (and recovery should we ever =
> need it) by investing in a NAS device to backup our database hot in the =
> wee morning hours using Recovery Manager and Legato Networker. Then =
> performing a file system backup to backup from the NAS to tape during =
> the day.
>
> Question for anyone using NAS devices in this fashion. Would disk =
> drives with SATA technology be "strong" enough to help me shorten our =
> backup window? I know that SCSI technology would be better but with our =
> budget constraints it's just at the top of our available budget dollars.
>
> Any experiences (positive or negative) out there from anyone using NAS =
> devices in this fashion would be appreciated.

Forward this to the list if you'd like to get feedback on this response... It is all brainstorming instead of direct experience with this setup.

So, basically you have a limited budget and you're looking for a way to optimize your backup speed, right?

What's the underlying issue being addressed? For example, are the backups interfering with performance? Are you mainly trying to reduce the amount of time needed to restore service after a failure? Are tape resources scarce and so you need to limit the time used? Maybe going over the SLA time limit?

What's your current performance and what do you want it to get to? For example, if you are currently shoving 50MB/s to the tapes, you are not going to get much performance increase if you are going to use gigE network connection to the NAS.

Okay, on to the fun part. From my understanding the main reason to go with a SCSI disk solution is for performance and durability. Industrial grade SCSI disks can take constant random access day-in and day-out where a consumer grade SATA will fall apart. SCSI also has more advanced disk controllers, especially over ATA, but SATA is starting to catch up.

Now, for a backup solution, you are basically going to stream sequential data to these disks for a relatively short period of time. I don't think you'll see much difference between SCSI and SATA performance, and the disks should be able to handle this amount of use no problem.

Here is where I think you can go crazy cheap. The data on this staging area is not unique in any way. You have your production system, obviously, and all backups are put immediately to tape. That means this staging area doesn't need to be high availability or massively redundant or anything expensive like that. Instead, buy two sets of (2 400GB SATA drives, 1 gigE cards, and 1 PCI SATA disk controller). Put those into any random PC sitting around. Hook one of 'em into a switch. Let the DB server and the tape server NFS mount the two drives as two mountpoints. Have RMAN use two disk channels, one for each disk. If you tune RMAN and the DB server can push the data out fast enough, you should be able to hit over 50MB/s. [Quick google check.] StorageReview has tests that show SATA drives starting at 65MB/s (outer edge) and ending at 37MB/s (inner edge). Okay, you might want to add a 3rd drive to guarantee network saturation. You can either add more drives and network cards to scale up, or if the PC's internal bus overloads, add more PCs as well (and more mount points).

Why two sets of hardware? If one blows up you just grab another PC from some random stranger's (manager's?) desk and *poof* you have another staging area. Management nervous about "single point of failure"? Have both of the PCs up and tell RMAN to duplex the channels to both of 'em. Or stick with the original plan, and then copy the files from one PC to the other.

Anyway, just some ideas.

Steven

--
http://www.freelists.org/webpage/oracle-l
Received on Tue Mar 15 2005 - 18:31:19 CST

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US