EVA4000 and 1TB fata drives

[inex]

We have a customer with EVA 4000 (xcs 6110) and 500GB fata drives with both iscsi mpx and fc hosts attached. The customer would like to swap out 14 of the disks with 1TB drives. Would this work?
(I see that XCS code needs to be upgraded)
Is it neccessary to upgrade the code on th mpx unit as weell?

-inex

[Víctor Cespón · Accepted solution]

1 TB FATA drives are compatible with XCS 6.110

XCS 6.110 is currently out of support so you should update anyway.

Yes, you can ungroup 500 GB FATA drives and replace them by 1 TB FATA drives

The space will not grow by 14x500 GB, the mathematics are more complex and depend on how the RSS groups end up.

Updating the firmware on the iSCSI router is always a good idea, but not related to this.

[inex]

The EVA4000 is fully populated with 56 drives.
The have single failure protection level. How can we calculate approx. what extra space we will get if we repalce 14 of the drives with 1TB disks?

-inex

[Uwe Zessin]

In the best case you win 13 raw TeraByte (protection space grows from 1TB to 2TB). That is true when each 1TB disk drive is 'married' (for VRAID-1 data) with another 1TB disk drive.

In the worst case you LOSE ONE TeraByte !!
That is true when each 1TB disk drive is 'married' with a 500GB disk drive. The protection space still grows from 1TB to 2TB.

[inex]

Not sure if I understood this 100%. We have only vraid-5 volumes and we will migrate out 14 x 500GB disk drives and install 14 x 1TB disk drives.
I thought maybe this would give brutto (14 x 1TB) - (14 x 500GB) - 1TB (ekstra for the spare)

-inex

[Uwe Zessin]

It does not matter what VRAID level you are using _now_. The low-level data structures are laid out for the extreme case - when ALL of the vdisks in a disk group use VRAID-1. You could change your mind and migrate all your vdisks to VRAID-1 any time.

[Víctor Cespón]

(14 x 1TB) - (14 x 500GB) - 1TB (extra for the spare)

No, it's not that simple. Internally the disks are divided in groups of 8 (RSS) and the RAID information is distributed among these sub-groups.

Typically each RSS will only have 1 or 2 1TB disks. The controllers will try to fill those disks with as much data as possible but without losing redundancy.
This will lead the the 1 TB disks not being filled as much as you expect, so the total avaiable space will be less than that formula.

[Uwe Zessin]

And the situation isn't even 'cast in stone'. If disks drives leave the group (e.g. due to failures or ungroups) or disk drives are added, the EVA can (depending on the internal state) reorganize those RSSes (split them, join them, move disks).

As far as I can tell, there is no optimization code in the controller firmware, so you can even LOSE FREE/PROTECTION capacity during such a reorganization if a large disk drive gets married to a smaller one.