Can NetWorker backup a SAN directly?

When you use Storage Node instead of "normal" LAN Backup the data is put over SAN to tape. So I think you have GB between your Fileserver and NSR -> 30MB/s. Just test fileserver as storage node.

The only way I know to bypass direct, is to do an RAW Device Backup of this SAN-Luns.
What OS have your NetWorker server?

Have you got the possibility to clone or snap this san luns on your storagesubsystem?
RAW backup by directive is also possible by 7.X.

All OS's are Windows 2003, all network is gigabit. SAN and tape hardware are all fibre connected to the same switches, tape drives are LTO3.

Also, the reason I think turning the file server into a storage node will not make much improvement is that other partitions on the same server backup much faster, and indeed even the 1Tb+ partitions occasionally "burst" to faster speeds, presumamby when it is backing up larger files rather than many small ones. Its quite noticable when this happens, there is a big jump in speed for a while, and then its back down to a trickle again.

And yes I know of using direct SCSI, but 1) This is (I think?) only available on 7.4, and 2) more importantly, we must keep indexes.

I know this whole thing is probably wishful thinking anyway, but I thought there was no harm in asking

From my experience the only way to keep the backup data within the SAN Fabric is to use DDS drives or else it hits the network. Metadata (indexes) will still travel over network to the server.

Our large windows file & print use DDS drives mapped to a UNIX hosted autochanger to get the throughput we need. Same sort of situations - LUNS up to 1.6TB & millions of files per LUN.

Rgds
Nick

For those WIndows File Servers with multi-million files, have you considered NetWorker SnapImage? It's way more efficient for those types of clients than regular file scanning because it bypasses the file system overhead by tracking changes at a raw level.

A raw backup is probably a bad idea, since you lose the ability to do a file level restore... restoration times would be awful. I'd try SnapImage first, it's intended for exactly this problem (assuming the problem is indeed trawling/indexing all the little files).

FYI, something like Avamar nodes might be a better fit, but SnapImage would be a lot less effort to deploy.

Hi all, thankyou for your replies.

It seems that there is no easy way out of this.

Doing a raw backup is easiest and the closest fit to what I want but does not provide file indexes, which are mandatory.

We have a lot of IBM FastT so we could use PowerSnap, but we don't have the capacity to be snaping LUNs (its not our hardware, so we can't just upgrade). Also, we are migrating to HP hardware now and there is no PowerSnap module for HP SANs that I know of, so we would have to get the SAN to do the snapshots. But at the end of the day, the trawling/indexing won't increase in speed, but there will be less impact since the LUN is not a production LUN.

SnapImage would probably work best, but it is not without problems either, such as the requirement to have its own driver for tracking disk changes (how much can you trust that?) and the fact that there is no incremental backup, only diff and full (I can understand why, since it is doing things on a block level, but it's still annoying). Anyone have any experience with SnapImage? Is it reliable?

We just bought a Netapp filer and I am trying to figure out how to backup the entire filer. From what I understand I can do this using NDMP, but have yet to understand how to write to a tape device. I only have one tape library and it appears that I need to specify the tape separately from the normal Networker backups since NDMP and normal Networker data cannot exist on the same tape because of the difference in how the data is written to tape.

I would use Avamar for those. SnapImage... well, I would not go for that to be honest, but you can try. Indeed, PowerSnap for HP HW does not exist. With so many small files you either wish to move this to NAS box and then use block level NDMP backup or some dedup solution which is where Avamar comes handy.

p.s. tkutil, looks like you missed the thread, didn't you?