One of the benefits of server virtualization is that hypervisor vendors provide IT professionals with a variety of VM protection options. For example, if an administrator wants to run virtual machines from a secondary data center in times of disaster, they could achieve that goal by building a geocluster or taking advantage of VM replication. But which approach works best?
VM protection option 1: Geocluster
Many IT professionals consider a geocluster, sometimes referred to as distance clustering, to be the best VM protection option. When properly configured, the virtualization software can immediately detect a data center-level outage and automatically transition running workloads to an alternate data center, where those workloads can keep working with almost no interruption in service.
Although a geocluster provides undeniably good VM protection, a number of issues must be considered:
A geocluster can be expensive and complicated to implement. Depending upon which vendor product you use, you may have to be careful about locating virtualization hosts in a way that prevents a quorum from being lost during a failover. Furthermore, some virtualization vendors impose strict distance and latency requirements. As a result, you may not be able to build a long-distance cluster if your data centers are located too far away from one another.
Building a geocluster requires a lot of hardware, and much of this hardware remains underutilized. If the host servers in the remote data center were heavily utilized, they would not have the available resources necessary to accommodate a large-scale failover.
VM protection option 2: Virtual machine replication
Virtual machine replication is another way to fail VMs over to an alternate location. This feature works by creating a standby copy of a VM on a remote host server. Depending upon the hypervisor used, it may be possible to establish two separate clusters -- one in the local data center and another in the remote data center -- and then replicate VMs from the primary host cluster to the secondary host cluster. This VM protection approach offers a high degree of protection, without the complexity of building a long-distance cluster.
Virtual machine replication also has its disadvantages:
Depending on the hypervisor, VM replication may not offer real-time protection. The replication process is typically asynchronous, which means synchronizations occur on a scheduled basis -- usually every few minutes -- rather than occurring in real time. Therefore, if a disaster were to strike your primary data center, you could lose a few minutes' worth of data.
VM replication typically results in the creation of an off-line copy of the replicated virtual machine. In the event of a disaster, this remote copy must be manually activated and brought online, depending on your hypervisor. In contrast, a properly configured failover cluster can provide an automatic failover from one data center to another.
Depending on the product, replication may not provide the required scalability. Microsoft Hyper-V, for example, requires replication to be enabled on a per-virtual hard disk basis. This isn't usually a big deal for organizations that need to protect a small number of VMs. But the only practical way to enable replication for large numbers of VMs is to script the operation through PowerShell.
Geoclusters and VM replication each have strengths and weaknesses, so IT professionals should review each technology carefully to determine which approach is best suited for their organization. While the major hypervisor vendors offer features that are similar to those offered by competitors, features such as clustering and replication come with a number of vendor-specific nuances. Once again, it is important to review these differences to see which VM protection feature is a viable option for your environment.
Options for protecting VM data: Three questions to ask
Snapshots might not be the best VM data protection
Software options for protecting virtual machines