HPE Storage Users Group

What would be the equivalent to a Hitachi shadow image copy? A full copy, physical copy, or virtual copy? I've never used 3par before.

I am looking for the best solution that will allow me the ability to mount a copy of my data to a backup server and allow resyncs with no or minimal impact to "parent" volume.

I currently back up approximately 90 terabytes of shadow image. I resync daily. Our largest database, approx 45TB averages 45,000 IOPS.... mostly reads.

Thanks,

If you want to maintain the current HDS methodology you would use a physical copy (clone) with a virtual copy (snap) attached to the source for periodic resync. To perform the backup after resync you'd then snap the clone copy and mount that snapshot for backup.

3PAR supports that method but it's typically only used for legacy compatibility and RFP tick boxes. In the 3PAR world a much better option would be to just take a virtual copy or several virtual copy snapshots, no need to resync, no impact on cache and minimal capacity overhead.

No or minimal impact to the source volume works differently on 3PAR vs HDS etc since everything is wide striped, source, clone or snapshot, so typically all share the same physical disks anyway. Counterintutively the clone resync will likely have the biggest impact on the source volume, either option is available to you but IMHO virtual copy would be the better and more flexible option.

Any idea what to do when a consistency groups needs to include more than 100 volumes?

Nope that's the limit today, but why would you need so many volumes in a consistency group ?
If you needed lots of small volumes on the previous HDS architecture to drive performance or simplify volume expansions then with 3PAR's system wide striping and virtualization you wouldn't need so many volumes to achieve the same. I'd assume there's a volume manager involved at the host layer to manage >100 volumes so maybe you could consider laying the data out more appropriately.

Have an oracle DB on HPUX.. approx 50TB most volumes are approx 250GB.

On HDS we've had good experience with keeping volumes no larger than 300GB for database data volumes

It's a common theme when coming from more traditional architectures especially big iron like HDS/VMax where getting the most out of the array required many many small volumes. Hence the need to support many thousands of volumes and huge consistency groups on those architectures.

3PAR wasn't built in that era or in that way and so doesn't require the same level of micro management around LDevs etc. Structuring the storage in that manner in many ways actively works against 3PARs architectural advantages, yes it'll work but you're going to be making a rod for your own back longer term.

You can still keep multiple volumes If you need to ensure very high queue depths etc but you could drastically reduce the numbers to simplify things whilst still maintaining performance. I'd take a look at the 3PAR architecture and concepts guide which should highlight a few of the fairly significant differences.

Why buy into 3PAR and treat it like a HDS/EMC box ?