Re: Raid Configurations and Oracle

We have a fairly large oracle database 15G on raid5, NCR 3000. We can back it up in one hour utilizing reel backup/librarian software and 1 tape drive / 6 concurrent streams.

Our tape drives are stk 9490 timberlines. They move very quickly. We get approx. 5M/sec utilizing streaming. Use of Raid 1 would definitely improve reading time. I think we can do much better with some tuning. If we were to use 2-3 drives we could backup 15G in approx 10-15 minutes. We will get around to tuning - nice luxury to have.

The stk silo's are the only real industry strength backup solutions on the market. If you are serious about backups I highly suggest getting one.

Problem with conventional unix backup methods like dd, etc it that they do not support streaming. Streaming is what really improves performance. Heck our tape drives can write fast than disks can transfer data.

Albert W. Dorrington wrote:
>
> In article <DyB1w2.K8_at_avenger.daytonoh.ncr.com>, David Jordan <David.T.Jordan_at_DallasTX.ncr.com> writes:
> |>
> |> ==========mthomas_at_digital.net, 9/23/96==========
 <snip>
> |>
> |> Mary,
> |>
> |> My experience comes from using an NCR 3550 system configured as Raid 5. Under
> |> this configuration, daily hot backups took 10+ hours to complete a 14GB
> |> database. As you can appreciate, this was intolerable. We reconfigured the
> |> system by going to Raid level 1 only, and the hot backup time dropped to 5
> |> hours and the users noticed a decrease in the length of queries.
> |>
> |> By going to Raid level 1, we cut our available Raid storage down by half, but
> |> it was worth it. There seems to be considerable controller overhead in
> |> running Raid 5. Consider this as you plan.
> |>
> |> In all fairness, though, please realize our backup methodology is not cutting
> |> edge. We are using the Unix dd command. Soon we will suppliment our backup
> |> strategy with new Digital Linear Tape Drives. Estimated backup times for a
> |> 20GB database is 3.17 hours.
> |>
> |> Best of wisdom,
> |> David Jordan
> |>
>
> A few comments on the above - you state you are doing a hot back up, but
> then say you are using the Unix dd command - don't quite understand
> how that could be.
>
> In any case - we are running a 20 GB database (two MTI/NPI-9100 RAID-5
> arrays with 10.2GB of usable space each, utilizing Seagate 2.1GB
> 7200RPM Baracuda drives.)
>
> The SCSI connection to the controllers is Fast Wide Single-ended and the
> disks themselves are standard SCSI bus.
>
> Our system's database transactions are mostly reads, but we experience
> quite a few writes per day as well.
>
> In our experience, our RAID 5 structure was FASTER than straight
> disks. (probably due to the Wide interface and faster disks.)
>
> I have seen benchmarks of the MTI RAID controllers against other RAID
> controllers and they seem to be faster than others on the market.
>
> As for your DLT comment - I definately agree with you, we have a DLT-4000
> tape drive and can back up 23 GB worth of diskspace in around 4 hours
> using 'dump' (an hour faster than our 4mm autoloader)
>
> - Al
> --
> Al Dorrington
> awdorrin_at_ictest.delcoelect.com Database Admin
> Delco Electronics - IC CIM Unix Sysadmin
> Kokomo, Indiana, USA Phone: 317.451.9655
 

-- 

Thanks,
-------------------------------------------------------
Joe Minnich