New DAE- Best practices for RAID setup

so today you are using this pool correct ?

are you planning to use this additional capacity to extend existing file systems or to create new ones ?

That's a good question. 

I was originally going to create new file systems.  In my VMWare hosts, filesystems are presented like this:

vm1-sas 1.41 TB

vm2-sas 924.73 GB

vm3-sas 2.76 TB

vm-unprotect-sata 592.37 GB (SATA-II)

backup 2.81 TB (SATA-II)

So basically I was going to create a vm4-sas and a vm5-sas and each would be around 2TB mark.

However your question raises an eyebrow.  I guess I could theoretically extend those filesystems.

But to keep it simple, lets just say I go with my original plan to create 2 new file systems to present to vmware hosts.

(This is NFS mount).

I also threw around the idea of more disks in a raid 5 group.  The thought was more spindles the better.  But then again its larger size so it would take longer to rebuild or replicate / backup / move / etc..

So the 2 ~ 2.5 TB filesystems are still manageable when presented to VMWARE.  Though it would be neat to see one giant file system and not think about where certain machines are stored (just pile them on there), I was always told not to be lazy and do something like that :-)

Yes that is correct.

There's clarsas_archive and clarata_archive but all of this will be added to clarsas_archive.

carsas_archive

Capacity 5.6 TB

Used 5.5 TB (97%)

Free: 190.5 GB (3%)

Used by

checkpoints for file system vm2-sas, checkpoints for file system vm3-sas, vm2-sas, vm3-sas, vm1-sas

Stipe size KB : 32

Number of Stripe Members: 4

See that's my original thought as well.  I was second guessing it when I was told that a 6 disk (5+1) raid-5 has more write penalty because it takes longer to calculate parity due to the extra disk????

Maybe this guy who told me that was mistaken?

Anyway if I keep it consistant..

The first 6 disks are RAID5 (5+1), present 3 equally divided LUNs off that, create a vm4-sas filesystem.

The second 5 disks are RAID5 (4+1), ''''', create vm5-sas filesystem.

The last disk is HOT SPARE LUN202 in this case.

Didn't know what you meant, but a quick search of Isilon shows me some cool stuff from EMC.

If money only grew on trees...

Wait, but would I still be here asking questions if my money grew on trees, rather than came from working in IT?  Eh, its still a fun hobby - just would be limited in hours.

Isilon sounds great.

Anyway the NX4 can't do RAID-10 can it?  I see an option for RAID1/0 but that slash in the name leads me to believe its not real raid 10.  More like striping / no mirror.

But when I pick RAID1/0 it says 8 disks are reccomended. (Options are 2,4,6,8,10,12).

Again its that slash in the name RAID1/0 that has me worried its not RAID10 (no slash between the 1 and 0).

sauerk wrote: I also threw around the idea of more disks in a raid 5 group.  The thought was more spindles the better.  But then again its larger size so it would take longer to rebuild or replicate / backup / move / etc.. So the 2 ~ 2.5 TB filesystems are still manageable when presented to VMWARE.  Though it would be neat to see one giant file system and not think about where certain machines are stored (just pile them on there), I was always told not to be lazy and do something like that :-)

Isilon ?

Also can a bad disk in DAE2 utilize a spare disk in DAE1 or 0?  Or would you reccomend having a hot spare in each drive enclosure?

Isilon - one huge file system, scales up to 15 Petabytes.

sauerk wrote: See that's my original thought as well.  I was second guessing it when I was told that a 6 disk (5+1) raid-5 has more write penalty because it takes longer to calculate parity due to the extra disk???? Maybe this guy who told me that was mistaken?

there is no extra parity calculation just because you have an extra disk, it might take slighly longer to rebuild to hot spare.

For File (NAS) usage we have always done 2-disk RAID1 on the storage processor and then let AVM on the data mover stripe across multiple LUNs from these

From what I remember the r1 AVM profile will try to find up to 4 LUNs on different R1 RG’s and stripe them togehter

i have never configured R1/0 on Celerra but is sure looks like it's treating it as RAID1 and not RAID1/0

hot spares are global, make sure you have at least one that matches your drive type and capacity.