Judging by some of the virtual machine (VM) management tools we evaluated recently, server backup software could soon be replaced by server replication tools.
The trouble with backups is they spend most of their life archived on some tape or disk storage system. Even if you test the tape as soon as you’ve made it, you can never be sure the backup will restore properly when you need it. In contrast, replicas are alive and well and ready to run on one of your virtual server farms.
The replication tools work by copying a running VM onto a secondary server. Then they monitor the VM for changes and from time to time they update the replica with the changes. The result is an up-to-date copy of the VM that’s ready to run. The idea is attractive enough as it stands, but it gains even more appeal when the secondary server is located in a distant city or on a different continent. Then it goes beyond the mere budgets associated with backup systems. It becomes a disaster recovery strategy.
Of course, this type of replication is something that can only be done in virtual server environments. It relies on a hypervisor to monitor the VM and snapshot its disk at an appropriate moment. Likewise, a replicated server is configured to use specific SCSI and network interface cards (NICs) etc., and these are unlikely to be available in your secondary server if it’s a physical rather than a virtual machine. VMs use virtual hardware – virtual NICs and virtual SCSI adapters – so it’s easy to ensure they are available on the secondary server.
As with many things in the world of virtual servers, it seems still to be early days for VM replication. Although there are currently a few replication products, they tend to go about things in different ways. For example, one uses data compression to minimise the amount of data at the network layer, while another leave network data uncompressed on the basis that this approach puts less load on the host server’s hardware. I prefer the second approach.
Both the products I’ve looked at required administrators to drive the failover process manually. But while one did some of the work for you, the other made administrators start the fail-over process from scratch. Surely such products will soon have buttons to trigger the fail-over automatically?
Meanwhile, both products monitored the VMs for changes and attempted to minimise the amount of data sent over the network by sending only the data that had changed. One product used the snapshot capability in the virtual machine’s hypervisor to actually track the changes. This product needs only to monitor the snapshot file and send that file when it reaches a certain size. The last time I looked, the other product compared the virtual disk files from the source and replica, which is relatively a slow process. However, the latest version of Vizioncore vReplicator offers the option of sending the snapshots in much the same way as Double-Take for Virtual Infrastructure. With only a couple of products, replication tools might sound like a whacky niche. But no doubt in a year or two there’ll be many more products to choose from. Server backup is dead, the replicas are coming.