Mailing List Archive

I also like to know if anyone has a strong opinion about this?

Nader


On Tue, 15 Oct 2002, Phil Hill wrote:

> Hi all,
>
> does anybody know, how fast is a R100 ?
> We want to build an environment, which reduce the downtime in
> case of a disaster.
> Following in mind, some Netapp Filers do a SnapMirror over a
> WAN-connection to a R100. Beside the R100 is a spare-system
> which could replace any production-filer. In case of a disaster
> we are able to copy the volumes to the spare-system and replace
> it with the broken filer.
> This should be much faster then a full restore from a local tape.
> I am talking about 0.5TB - 1.5TB per remote filer.
> We also would do only a tape-backup from the R100, no local tape
> drives in the remote offices.
>
> Questions:
> - What is the real read-performance of a R100?
> - What is the fastest way to duplicate the broken volumes from the
> R100 to the spare-system? volcopy ? ndmpcopy ? something else ?
> - How long does it take to copy a 1TB volume to another filer.
> - Does anybody have something like above described ?
>
> Thanks in advance
> Phil
>

well, and R100 uses the same head as an 880. if you carefully build
the volumes and raid-groups such that you spread the I/O load across
SCSI busses and ATA-SCSI bridges, you can make the R100 go pretty fast.

in my testing, the sustained seauential read performance is 120 Mbyte/sec.
for mixed file sizes, I suspect it to be somewhat less. writes into the
R100 sustain 60Mbyte/sec, with gusts up to 65 Mbyte/sec.

I rarely move whole volumes between filers, so I don't know the best
way to do that. suggestions welcome.

-skottie




Phil Hill wrote:
> Hi all,
>
> does anybody know, how fast is a R100 ?
> We want to build an environment, which reduce the downtime in
> case of a disaster.
> Following in mind, some Netapp Filers do a SnapMirror over a
> WAN-connection to a R100. Beside the R100 is a spare-system
> which could replace any production-filer. In case of a disaster
> we are able to copy the volumes to the spare-system and replace
> it with the broken filer.
> This should be much faster then a full restore from a local tape.
> I am talking about 0.5TB - 1.5TB per remote filer.
> We also would do only a tape-backup from the R100, no local tape
> drives in the remote offices.
>
> Questions:
> - What is the real read-performance of a R100?
> - What is the fastest way to duplicate the broken volumes from the
> R100 to the spare-system? volcopy ? ndmpcopy ? something else ?
> - How long does it take to copy a 1TB volume to another filer.
> - Does anybody have something like above described ?
>
> Thanks in advance
> Phil
>

--

Scott Miller | Animation Technology
work: skottie@dreamworks.com | Dreamworks Feature Animation
life: skottie@pobox.com

> We also would do only a tape-backup from the R100, no local tape
> drives in the remote offices.

Phil,

I recommend you get an R100 loaner from Netapp and attach your tape
drives. Depending on your drives, the R100 will not be able to properly
keep your drives running at full speed.

For example, AIT-3 drives connected to host running NFS to R100 or F880
can do 20+MB/sec. AIT-3 drives connected to R100 or F880 directly can
only do 10+MB/sec. Also, watch CPU utilization when you run NDMP locally
vs. backups over NFS. Each drive will use 30% of R100 CPU, but only 15%
of F880 CPU (even though the CPUs are the same).

You truly need to test this in your environment before making purchases.
If you're backing up terabytes of data from remote offices, get a true
F880c or F960c instead of R100. If you have a standby filer sitting
there, you could move the shelves from the snapmirror destination filer to
the standby filer with no data copying. With the R100, this is
impossible.

/Brian/
--
Brian Long | | |
Americas IT Hosting Sys Admin | .|||. .|||.
Phone: (919) 392-7363 | ..:|||||||:...:|||||||:..
Pager: (888) 651-2015 | C i s c o S y s t e m s

The R100 can be very performant, if you build you volumes correct. We find
almost the same speed as local attached sun T3 disks, when we do large
writes and reads (dumps of our sybase).



We use the R100 for our primary storage of us PACS-system (digital X-rays).
We produce more than 20 GB / day. 30 workstations read the pictures and do
the diagnostics. It works perfect.



We do also snapmirror some filer volumes to the R100. When there is a
disaster, I will serve the data directly from the R100. I'm convinced that
he (or she?) can do that. I'm just talking about "fileserver" stuff. No
databases.



So Phil, what kind of data do you have. When it are just "files", why copy
the data to your backup-system? When it is a database or very write/read
intensive application, ok, I think you must copy your data.



I would use the (qtree) snapmirror. But it takes time, but while you are
waiting, you can try to serve the data from the R100.



I hope this will help,



Reinoud Reynders


University Hospital Leuven
Belgium


----- Original Message -----
From: "Phil Hill" <phill@ultra.de>
To: <>
Sent: Tuesday, October 15, 2002 4:58 PM
Subject: R100-Filer disaster scenario


> Hi all,
>
> does anybody know, how fast is a R100 ?
> We want to build an environment, which reduce the downtime in
> case of a disaster.
> Following in mind, some Netapp Filers do a SnapMirror over a
> WAN-connection to a R100. Beside the R100 is a spare-system
> which could replace any production-filer. In case of a disaster
> we are able to copy the volumes to the spare-system and replace
> it with the broken filer.
> This should be much faster then a full restore from a local tape.
> I am talking about 0.5TB - 1.5TB per remote filer.
> We also would do only a tape-backup from the R100, no local tape
> drives in the remote offices.
>
> Questions:
> - What is the real read-performance of a R100?
> - What is the fastest way to duplicate the broken volumes from the
> R100 to the spare-system? volcopy ? ndmpcopy ? something else ?
> - How long does it take to copy a 1TB volume to another filer.
> - Does anybody have something like above described ?
>
> Thanks in advance
> Phil
>