How much disk space do I get with a Netapp 2040?

[modestyblaise]

We have a  new Netapp 2040 with 12 internal 300GB SAS disks and 2 heads configured as Active/Passive. It has been configured as follows;

2 Parity disks

1 Hotspare

3 System disks for the Passive head

This leaves us with 6 disks for useful data storage, and a total of 900 IOPS according to the consultant who installed it. We expected more than this. Does this seem reasonable, and is there some other way to do the configuration?

[mheimberg]

Each head or controller owns a set of disk, including spare(s).

When you write

2 Parity disks
1 Hotspare
3 System disks for the Passive head

and 6 unused disks this does not make sense to me.

So the layout should be:

controller/head 1 (active)

2 Parity disks

5 Data Disks

building the root-aggregate, containing the root-volume vol0 and data

1 Spare

controller/head 2 (passive)

2 Parity disks

1 Data disk

building the root-aggregate, containing the root volume vol0

1 Spare

I have also configured passive system with RAID4 instead of DP, giving one more disk as data disk for the active controller.

In the Storage FAQ you can find:

In an HA pair configuration NetApp recommends the following to maximize storage utilization:

 Controller 1 (RAID-DP):

o

Number of Aggregates = 1

o

RAID Group Size = 8

Number of Data Drives = 6

Number of Parity Drives = 2

Number of Hot Spares = 1

 Controller 2 (RAID-DP):

o

Number of Aggregates = 1

RAID Group Size = 3

Number of Data Drives = 1

Number of Parity Drives = 2

Number of Hot Spares = 0*

*As a result of assigning no spare drives to controller 2, warning messages will be generated that can be safely ignored (

raid.min_spare_count can be set to 0 to avoid warning messages).

so now we got the disk-layout.

How much do you get out of it, is the question. And the answer is as always: it depends.

There is a rule of thumb, stating that you get roughly 900-1200 IOPS out of 6 SAS disks. But what is the workload? Do you read 4kB or 64kB per IO? Which protocol (eg. CIFS has a lot more overhead than NFS)?

So first of all one should know your workload (protocols, IOPS, bandwidth, #clients etc) - or simply: what are you going to do? And what do you expect?

BTW there are lots of sizing tools from NetApp your consultant surely knows.

Mark

[brendanheading]

The consultant has set it up with by-the-book NetApp best practice, but it's probably not giving you the best value.

If you configured the passive controller with RAID4 and no hot spare, that would leave 10 disks for the active. Two RAID-DP and one hot spare there gives you 7 data disks. You should end up with about 1.7TB right-sized. Are they 15000rpm disks ? If so you should get another ~180IOPS based on the usual rule of thumb.

Seems like a waste to have a dual-controller 2040 with only 12 disks.