Synthetic or Pure Disk - Looking for ideas

[Mathew_Burke]

I want to use synthetic backups or pure disk so I can centralize the backups of our remote divisions (over 10 Mbit connections for the most part).

 

So I am testing Synthetic and can create full synthetic backups to tape. But the thing is I want a method to restore servers without pulling it over the WAN from my disk-based backups because a WAN based restore would be too slow.

 

We can use the LTO tapes (created synthetically) but in the scenario there would be no media server in the remote division to do the restore, and I don’t want to lug a big heavy server with me when I go out to restore a crashed server.  So a LTO device to run off a laptop would be nice, or even a USB hard drive (with a DSU) would be better.

 

The problem with a USB external HD is Veritas wants to connect that to a particular media server (i.e. not my laptop).

 Any ideas?

[RAPHAEL_F_CARVA]

Puredisk is probably the best option cause while it reduces the traffic on WAN, the data is stored on disk (can be duplicated to tape by NetBackup integration) and can be mounted or accessed as CIFS partition so, even if you don't want to mount the share remotely (in your remote office's server) in case of data loss, you can still open the partition * on your notebook locally, copy the data you will need to your USB external disk and carry it to your remote site. 

* data is de-duplicated, encrypted and protected on puredisk server but with the right password to access, it's transparent, you see and can copy the data as normal data. It's very cool.

[FED_SE]

Another Idea maybe to cluster the NBU server as NBU supports this architecture. This way you can have a primary data center in full production while in the back ground send catalog DB to DR site and sync the catalog with the remote cluster node. Then there is the need to get the data there. Puredisk is an excellent way to accomplish this as it has replication built into it. With regards to licensing you are only charged for the licensing of the front-end terabyte. Additional copies of the deduped data cost nothing other then some hardware at the other end.

Another thought maybe to look at Storage foundation HA/DR for clustering and Volume replication. 

If the critical servers are critical enough to require quick SLA turn around, it might be a good plan to start looking at SF VCS and VVR for recovery. 

[Mathew_Burke]

Thanks for the replies, it is good to get ideas from another perspective.