I work in a small company which is getting bigger all the time. We have out grown our old backup system (a Small NAS box and Sugar Sync) and would like to move to something better...
We currently have 3 servers, 2 Win2k3 boxes and a 2k8 box. One of our servers is running SVN with all our code on it, and this is the most important machine to get backed up. We also have SQL boxes, Oracle instances and MySQL installed too...
I have been looking at offsite backup plans, and have been thinking about the following:
- Take all the machines we currently have and Virtualize them using the P2V tools in System Center Virtual Machine Manager.
- Have the VHDs stored on a Nexenta or Solaris machine using ZFS and iSCSI.
- Using ZFS's snapshot tools, we can take a snapshot of the instances while they are running and back them up to Amazon S3 or similar. then just backup the changes between nights.
- if a machine fails, just replace the physical box and add it to the HyperV pool. copy the VMs on (copy is not the right word, given the files are stored on iSCSI, but hopefully you know what i mean).
- As long as the SAN is built correctly, we should be ok for a disk failure (ZRAID or ZRAID2)
- since everything is backed up to S3, if we loose the office (Fire, Meteor Strike, Aliens, etc) we can get our data back (as long as Amazon still exists).
What do you think? Is this a feasible solution?
PS: advantage i just though of using ZFS: Data Deduplication should (in theory) mean we store less on the iSCSI box. If we upgrade all our machines to 2k8R2, we only need to store one real copy of it... the rest are de-duplicated...
Take a look at MS Data Protection Manager if you're already using SC and planning on Hyper-V. It's a pretty good backup product. I personally run this and a secondary offsite copy of it to backup my primary DPM server. It works pretty well if you've got a heavy investment in Microsoft already.
Taking a SAN snapshot of a running VM particularly with a DB running can be questionable at best. Without quiescing the VM you'd run the risk of trashing your DB. Basically the snapshot would be as if you pulled the plug on the server. It might work but it might not. It would depend on your requirements whether that would be acceptable or not.
We are a small company with a small VMware ESXi cluster (2 hypervisors with about 15 VM's, both Windows and Linux) and use Veeam Backup & Replication to backup all virtual machines. We currently backup daily to a large local disk and copy that backup file to tape weekly for off-site. Another supported method is to write directly to a remote Veeam server (either your own or from a provider). Recently Veeam has also added 'cloud'-backup so you can backup directly to various cloud storage, like S3, Azure, etc.
We're pretty happy with it. Efficient, compressed, de-duplicated and fast. There are multiple levels of recovery possible, such as full-VM, virtual disk, individual file or even individual records (for Exchange and AD). For fast recovery one can boot the VM directly from the backup storage (with reduced I/O performance) and move to production storage later. A downside is that Veeam's limited to virtual machines. VMware and HyperV are supported, but no physical servers.