save device configuration

Hello PGK and Happy New Year. One question? Shall those snaps devices be mapped to FAs all the time? Shall those snaps devices be masked to a mount host? Especially if the snaps are mapped from the beginning then you have to use FAs ports that are not utilized so much. If it is possible try to create a separate disk group for the physical drives and try those physical disks to be FC drives. If you are in DMX3 or DMX4 create the SAVE devices with 65520 cyls. Regarding RAID type it is better to create the SAVE devices as RAID1 and you will have better performance.

Quincy, THANK YOU VERY MUCH.

The VP Snap (really clones) can use the standard thin pool devices (TDATs) so there is no reason to make a seperate pool of devices.  You can have your source and target using the same VP pool.

how does this recommendation change with the new snapshot technology in 76 ?

Hi Quincy,

thank you for the information , we have 136(450GB) FC drives available and we need about 10TB of snap save space, i calculated  taking 30% change of total storage allocated to sap servers ie 26TB.currently  the default pool  is raid5(3+1) and it has 40*8GB save devices.since our FC thinpool is oversubscribed we donot want to use all FC disks for snaps.

would it be good idea to create a seperate diskgroup with 32*450 FC disks giving us 432*.75 =10368GB for save pool and can we create 75GB save devices and add it to existing pool?

Thanks in advance

GK

If you already have VP (thin pool) technology, you would be MUCH better off to simply use VP Snaps (space saving Clones) instead of the older traditional SNAPs which use SAV/VDEV devices.  With VP snaps you don't have to have a seperate pool for your target devices.

If you insist on using SAV SNAPs, then U would add those 32 FC drives to your existing VP pool and then delete one TDAT from every drive and then change those to SAV devices.  32 disks isn't enough for a SAV pool.

we are at microcode version 5875.269.201. i thought  you said we can use VP snaps only if we are at 5876 code!

why would i want to add 32 Fc drives to existing VP pool and delete one TDAT from every drive and change to save devices if i can just create save devices directly from disk group from what ever space left after creating TDATs.

Yes, but 76 code will be the target code level soon.  Seems a lot of work when the new feature is just around the corner.

I thought you had a pool already in production.  If not, then just make one hyper of SAV on every FC drive for your pool.  Making them the same size as your TDATs may be good if you want to later go to VP SNAPs and turn them in to regular TDATs.

Hi Zx48k,

first of all i agreed with Quincy's last comment. It is the ideal scenario that you can do regarding Snaps and VP Snaps at 5876 microcode.

Also, because i had implemented my Snaps on DMX4 on 5773 microcode and now at VMAX on 5875, i have created 8 hypers/slices per disk on 300GB 15K FC Disk.

Of course the BE controllers, the TDATs and physical disks where the SNAPS are reside are more or less 80% utilized, but does not create any performance issues to the box or to the others applications.

This is point of view, but for sure i would follow Quincy's advise if i had 5876 on the VMAX. 

SAV are different than TDATs.  We found that 1 hyper per disk gives better performance.  You do NOT want the SAV devices to get high WP counts, and they receive 100% random writes.  This is why you want the SAV spread EVENLY spread over as many spinning disks as possible.

Good morning quincy and Zx48k,

sorry for my mistake, i meant SAVE devices and NOT TDATs in my reply.

Sorry again.

In any case, at least here in Greece Quincy, due to financial crisis, they don't pay so much attention to performance regarding the Snap Pools and the customers buying a small amount of disks.

For example:

I have a customer that they have bought only 30 FC disks, in order to create SAVE devices for Snap Pools added them in a specific disk group.

The applications are Billing and ERP and the changes commited in those pools are pretty much a respectable amount of changes, reaching some times and the 99% during the working hours..

Thus has as a result, the physicals disks to be always RED.

We proposed to them, more or less, your scenario but no money no honey.

Again, it is not that difficult to take the 30 drives, add them to the current pool, then remove 1 TDAT from every drive.

If the SAV pool becomes a write bottleneck, and it receives 100% random writes when activated, the entire system will suffer.

Hi Quincy,

thanks for the advise. I will try to do your suggeston.

If i convice the customer to buy more physical disks for the specific disk group, and if i follow your suggestion on the newly installed disks, shall i face any issue (performance) on the Snap Pool?

Quincy thank you so much.

Your info it is more than appreciated!!!!!!