Congratulations on the acceptance of your proposal.
I don't know what your company's policies are in regard to data retention and archiving so it would be hard to comment on what solution would fit. In general though, there are a few options you can consider:
1. Use of disk class backup under current backup software
Most of the backup software today lets you create a tiered backup approach by identifying different classes of backup devices. This lets you allocate local SAN disks as primary backup media and tape as secondary media. This tiered backup approach allows you to reduce your backup window by first moving data to fast disk then, archive to tape at a later time. Local disk backup is a great way to reuse all those disk shelves with old disks in them. The old disks are most likely fully depreciated but instead of throwing out that investment you can reuse it as inexpensive temporary backup storage. If they are SCSI disks, you can connect them to the SAN with a data-router or connect them to a server cluster and share them out as file share, which brings us to the second approach.
2. Backup to NAS
There are a number of vendors who provide NAS appliances that can be used to store backups. Most applications require fast block access so this is where SAN-based block access comes in. NAS is file based and uses CIFS and NFS for access. But if you have a fast network, say Gig-E, you can use NAS storage as a backup repository. The NAS can be local or even remote which would solve any off-site archive requirements. You can create your own highly available NAS appliance by using clustering software to share out host-based file shares. The cool thing with this approach is the ability to reuse older cheaper disks as the pool of storage for the shares. If you have a server with a bunch of free PCI slots, you can add SCSI adapters to connect to old SCSI disk shelves and use Y cables with end terminators to share those disks between clustered servers. (You would be creating a SCSI cluster.)
3. Use off-site mirrors
One cheap method that works is to create host based mirror volumes that span geographic locations. This does not require any cluster software or even backup software since everything is done through the operating system or file system software. The ability to create a "wide area mirrorset" is dependent on three things:
a) The performance requirements of the application
b) The bandwidth of your network
c) The availability of iSCSI connections
To create a mirrorset over IP would require block-based access to the disk on the remote side. iSCSI can provide this over your existing IP network. You can use iSCSI agents on the host for connection to remote iSCSI resources or you can use TOE (TCP/IP offload engine) capable network adapters that will speed things up a bit. I suggest a minimum of dedicated 100base-T network speed to the remote location. You can get iSCSI gear from companies like Nishan, Cisco, etc. to provide the connection to the remote FC SAN.
Click for Part 2
This was first published in December 2002