Re: 306GB drives!

From: Mike Naime (mnaime_at_kc.rr.com)
Date: 08/16/03 

Date: Sat, 16 Aug 2003 05:32:43 GMT

OK Rob, I'll bite on this discussion.

Rob Young <young_r@encompasserve.org> wrote in message
news:Wea6H8AO1F0F@eisner.encompasserve.org...
> In article <5uYZa.78482$7O4.1900368@twister.rdc-kc.rr.com>, "Mike Naime"
<mnaime@kc.rr.com> writes:
> > According to our sales rep, we should see 306GB EMA style disks shipping
> > sometime in November.
> >
> > So, for raw space, thats 84x306 = 25704GB. 25.7 Terrabytes of data.
> > So, If I populate a T-5 configuration with those (4 HSG's with 84 drives
> > each), I can get 102.8TB of data in 5 feet of data center row space.
> >
> > That's the same amount of space that I have in my existing 36 HSG
cabinets
> > populated with 18, 36, 72, and a few 144 GB disks.
> >
> > I just hope that the 8.7 VCS cards supports that size of drive.
> >
>
> Why or how would you use them?

BACKUPS!!!! BACKUPS!!!! BACKUPS!!!!!BACKUPS!!!! BACKUPS!!!!
BACKUPS!!!!!BACKUPS!!!! BACKUPS!!!! BACKUPS!!!!!BACKUPS!!!! BACKUPS!!!!
BACKUPS!!!!!BACKUPS!!!! BACKUPS!!!! BACKUPS!!!!!BACKUPS!!!! BACKUPS!!!!
BACKUPS!!!!!BACKUPS!!!! BACKUPS!!!! Temp Space.BACKUPS!!!!!BACKUPS!!!!
BACKUPS!!!! BACKUPS!!!!!BACKUPS!!!! BACKUPS!!!! BACKUPS!!!!!BACKUPS!!!!
BACKUPS!!!! BACKUPS!!!!!BACKUPS!!!! BACKUPS!!!! BACKUPS!!!!!BACKUPS!!!!
BACKUPS!!!! BACKUPS!!!!!BACKUPS!!!! BACKUPS!!!! BACKUPS!!!!!BACKUPS!!!!
BACKUPS!!!! BACKUPS!!!!!BACKUPS!!!! BACKUPS!!!! BACKUPS!!!!!BACKUPS!!!!
BACKUPS!!!! BACKUPS!!!!!BACKUPS!!!!
DB export from PROD to non-prod.BACKUPS!!!! BACKUPS!!!!!BACKUPS!!!!
BACKUPS!!!! BACKUPS!!!!!BACKUPS!!!! BACKUPS!!!! BACKUPS!!!!!BACKUPS!!!!
BACKUPS!!!! BACKUPS!!!!!BACKUPS!!!!!BACKUPS!!!! BACKUPS!!!!
BACKUPS!!!!!BACKUPS!!!!!BACKUPS!!!! BACKUPS!!!! BACKUPS!!!!!
More temp space.

Oh, and did I mention backups?

>
> Would you put anything mission critical on a 5 member
> RAID-5 (for example) and serve that LUN up to VMS or NT?
>

NT - NO. Citrix servers can be rebuilt easily. Not worth putting that on
the SAN.

W2K clusters - Yes, We do it. They are using about 3% of my SAN spindle
count. The Winders world doesn't like to play nice on the SAN. They are in
the process of moving off the SAN to MSA1000's

AIX on a COMPAQ SAN. about 11%. Hey... Who let them in here. :-) You
let one in the door and look what happens.

VMS Yes! 85% of SAN space. We have one Production client system that has
an oracle database on 4 different 6x18GB raidsets.

Everyone was po-pooing the raidset and no local drives when we set up our
data center. Our Oracle DBA's told us that we needed bazillions of mount
points. The other VMS folks told us that we had to have local drive for
page and swap files. We do not have any noticable sustained disk queues.
You see some during the backup windows, but not the rest of the day.

When the Compaq/HP performance folks came out to show us how to use the T4
data, they commented that it was a welcome change to see a bunch of VMS
systems that did not have any real performance issues to speak of. As a
matter of fact, our spikes on CPU/disk where shown to be directly caused by
all of the monitoring/performance collectors! On some clusters, these where
the processes that used the most overall CPU and disk IO.

> The reason I ask ... is an old favorite subject of mine
> has come up:
>
>
http://groups.google.com/groups?selm=Lsvhk3%24YxMND%40eisner.encompasserve.o
rg&oe=UTF-8&output=gplain
>

Personally, I think that shadowing is one of the biggest performance hits
that you can take on a VMS system. Take the redundancy off the VMS level
and put it in the storage controller. This free's up a lot of the system
resources. Also, when you have a system crash. No 6+ hour performance hit
from all the shadow copy merges. One of the BIG complaints heard from some
of our remote customers with LARGE databases.

The EVA really is making it potentially possible to do away with shadowing.
You do your backups and redundancy at the disk controller level. Not at the
OS level.

> ---
>
> From: Rob Young (young_r@encompasserve.org)
> Subject: Re: disk benchmarking
> Newsgroups: comp.os.vms
> Date: 2001-05-29 10:23:36 PST
>
> What happens when that RAID5 of 4 - 500 GByte drives takes 2 days to
> restore (5 days?), etc. etc. That probably won't happen, so where
> is this headed/used (large drives)?
>
BACKUPS! I would say that at least 1/2 of my SAN space is now dedicated to
backups.

How many copies of your mission critical data do YOU have on disk Rob??? Or
is it all going to tape?
Tape is TOO SLOW for restores! You had better have at least one copy online
to restore from.

Guess what. Fairly soon you will not be able to buy those small disk drives
anymore.
Have you tried to purchase any 18GB disks recently????
HP has already phased out the 10K 18GB drive. You had better hoard them.

> ---
>
> The point is with a 5 member RAID5 and HSG (using those 300 GByte
> drives) - you have a 1.2 TByte LUN. When/if the RAID5 array blew
> out ( didn't rebuild properly) - you have a 1.2 TByte LUN to
> restore.
>
> So you wouldn't use RAID5. Okay, a 360 GByte mirror set blows
> out. You get the picture.

It all depends on what your management is willing to risk/pay for.
If I could convince my management to double my HSG count, I would not need
to have raidsets. :-)

In 3 years, we have had less than 1% failure rate in our disks.
Auto-sparing has made a drive failure a non-event.

Anyone out there actually have an HSG80 raidset that did not rebuild
correctly on a drive failure?
I have lost an entire channel twice in the BA-370 models. Not yet in the
EMA's.

>
> EVA and others get around this by partitioning - whether true
> virtualization at the hardware level like EVA does - or other
> clunkier methods of competitors.

Our EVA (Populated with 146GB drives) is used to server up LARGE backup
drives so that we can SNAP and send to tape.

>
> I wouldn't stick 360 GByte drives in an HSG80,

Actually, it is a 309GB drive. An odd number. I did not remember correctly
when I posted earlier.

> unless of course you could take 5 , 10, 20, 30 hour downtimes
> for restoration - worst case of course (pick a worst case for
> you - does it ever get less than 5 hours blowing out and restoring
> a modest 360 GByte mirrorset - 20 MByte/sec restore speed?).
>

Worst case is that I restore the raidset from the backup drive. If it is a
backup drive that goes bad, what do you do? Restore it? or overwrite it?

We have seen 35GB/hour average transfer rates on disk-disk in the HSG's.
75GB/hour going from HSG to EVA. It all comes down to the question of
How many drives you need to restore?
How fast you can push it from one SAN drive to another SAN drive.
How many systems you can have transfering it in parralel? One per drive if
you are doing an image backup. More if you can split the data up.

Mike

Relevant Pages

  • Re: Computer problem, need help
    ... defining the three levels of backup and recovery. ... has been corrupted, use backup and restore. ... hard drives and a reliable backup utility like Ghost. ...
    (soc.retirement)
  • Re: Computer problem, need help
    ... Having a backup is often the only solution. ... system restore for software failures ... I found the only one to be 100% reliable was Ghost 2003. ... has problems with SATA drives and perhaps problems with partitioned ...
    (soc.retirement)
  • Re: XP System Backup. ASR No floppy drive
    ... My own choice would be to install a fresh copy of Windows and then run restore using that. ... hoping to use System Tool Backup to create ASR. ... but says Invalid or damaged bootable partition when I ... The KB actually discusses which USB CD drives that the XP ...
    (microsoft.public.windowsxp.setup_deployment)
  • Re: RAID, SCHMAID.......
    ... And risky to change it later without a full image backup. ... partition, & then restore only the OS, but still verydoable. ... partition before an install of something you havent installed before, ... but so many ppl make a mess of things esp with today's big drives. ...
    (comp.sys.ibm.pc.hardware.storage)
  • Re: Computer problem, need help
    ... >>> defining the three levels of backup and recovery. ... >>> has been corrupted, use backup and restore. ... >>hard drives and a reliable backup utility like Ghost. ... doesn't work I use my preferred backup utility, ...
    (soc.retirement)