ONTAP Hardware

8-8 Port Lun Group

BILL_BLAIR
3,965 Views
  1. Is this supported in 8.1.2 CMode?  Or only 8.2 CMode?
  2. And is 8:8 with a single lun group an advantage over 4:4 with 2 lun groups?
  3. And how are luns chosen for multiple lun groups within the array?
  4. What is the best practice for aggregate creation when there are multiple lun groups?

Thank you,

Bill

1 ACCEPTED SOLUTION

isaacs
3,965 Views

Hi Bill,

1.  Beginning in 8.1 cDOT

2.  8-8 is beneficial only with very large LG's, or with very fast disks.  Can the disk provide more than 4gbs of throughput?  Then 8-8 might help.  Do you have 200+ LUNs in the LUN group (max raised to 254 in DOT8.2)? Then 8-8 might help.

3.  Can you clarify?  I don't understand the question.

4.  If you had 2 LGs, then generally half the LUNs from LG1, and half from  LG2, would make up a single aggr.  So that the aggr is spread across  both groups of LUNs.

Dan Isaacs

View solution in original post

4 REPLIES 4

isaacs
3,966 Views

Hi Bill,

1.  Beginning in 8.1 cDOT

2.  8-8 is beneficial only with very large LG's, or with very fast disks.  Can the disk provide more than 4gbs of throughput?  Then 8-8 might help.  Do you have 200+ LUNs in the LUN group (max raised to 254 in DOT8.2)? Then 8-8 might help.

3.  Can you clarify?  I don't understand the question.

4.  If you had 2 LGs, then generally half the LUNs from LG1, and half from  LG2, would make up a single aggr.  So that the aggr is spread across  both groups of LUNs.

Dan Isaacs

BILL_BLAIR
3,965 Views

Hey Daniel,

Thanks for the explanation on 8:8 configuration.  Where I'm confused is how to recommend they layout the disks on the arrays.  And I'm confused on how luns are spread out across lun groups.

The customer has an AMS 2500 with these disks:

32 x SAS 15K 300GB

243 x SAS 15K 600GB

133 x SAS 7K 2TB

It will be attached to a 6240 cluster mode system mostly serving CIFS in a data analytics environment.  Random Reads.  If we utilize all raid 6:

  • Luns are created from each different disk type.
  • From you answer above I assume we can group luns from each different type into two lun groups.  i.e.  luns fro all 3 disk types are spread over the two lun groups.
  • Create aggregates across both lun groups as long as each aggregate contain luns from one one disk type.
  • Best aggregate RG size = ?.

Correct?

I appreciate your help.

isaacs
3,965 Views

OK, this explains why I gave you a wrong answer in the other thread.  I think I meant to reply to this message. 

Will you be using virtual pools on the AMS?  Or traditional RAID groups?

If traditional, then 5+1 RAID5  on the 300g give you 5 RGs, ~1.5TB each.

For the 600g drives, 5+1 R5 gives you 40 RGs of ~3TB each.


For the SAS drives, I would create ~1.5TB LUNs  (1 LUN per 300G RG, 2 LUNs per 600g RG)  That would give you 85 LUNs to split between at least two aggrs.  Note that to avoid spindle contention, no aggr should have more than one LUN from any given RG.

For the 7K drives, I'm guessing 6+2 RAID6, but would defer to the storage admin.  I'm going to assume these disks will not have performance as a foremost requirement. I would create 4 LUNs per RG @ ~3TB each  (16RGs, 64 LUNs).  Ideally making 4 aggrs.  But if performance doesn't matter, then fewer aggs would probably be OK.

BILL_BLAIR
3,965 Views

Thanks Daniel.  This helps tremendously.

Public