Sergey Galushko - Fotolia


The relationship between VM clustering and high availability

Server failure can cause devastating setbacks and cost businesses untold amounts each year. Avoid the dangers of disaster recovery with VM clusters and high availability.

As IT administrators know, VM clustering and high availability are intrinsically linked. Clusters -- groups of two or more servers that function like a singular system -- improve flexibility by allowing workloads to be moved among other servers in the cluster. In the event of a server failure, VMs are restarted on another server in the cluster, thus enabling highly available workloads.

While the correlation between VM clustering and high availability (HA) may be easy enough to understand, it can be more difficult to implement. Creating HA clusters can be time consuming, as selecting the number of VMs to include in your cluster can take a bit of guesswork, and comes with a variety of challenges. Fortunately, we're here to help simplify the process -- check out these five quick tips to ensuring, tailoring and securing high availability with VM clustering.

Finding the right balance for VM high availability

Although virtualization has irrefutably improved server consolidation and increased workload provisioning and migration flexibility, it's far from perfect. Hosting more workloads on fewer physical servers can mean large-scale outages in the event of a hardware failure. The best way to combat the vulnerabilities of virtualization is to ensure that all elements of your deployment, including VMs and hypervisors, are resilient and dependable. This is best accomplished by testing out a mix of software and hardware options -- including VM clustering, hot spares, snapshots and even multiple Ethernet ports -- to strike the right balance and increase workload availability.

How to refine host and VM availability in vSphere

In vSphere environments, HA is crucial to uptime; using VMware vSphere High Availability settings, users can tailor their HA to redistribute workloads after ESXi host crashes as well as VM operating system and application failure. This is accomplished by "heartbeats" -- communication pings sent from a guest VM to VMware HA. If the VM in question does not meet the necessary HA requirements, VMware HA will reset it. VMware HA requires a host cluster, which is defined as two or more ESXi hosts using shared storage.

VM clustering
Using Microsoft Cluster Services for virtual machine clustering

Are pods for virtualization worth the clustering headache?

Converged infrastructure pods have earned a reputation as a good option for virtualization, in large part due to their consistency and supportability. However, this doesn't necessarily mean they're the right choice for every virtual infrastructure, particularly those dependent on HA. Not all pods provide full redundancy and, as such, are liable to become a single point of failure. This is especially troublesome if a pod is employed as a self-contained host cluster, as it prevents VMs from conducting failover to other hardware. While this shouldn't discourage users from using pods to run virtualization hosts entirely, it's important to take into consideration in order to best protect your data center.

VMware vSphere clustering requirements and challenges

Looking to configure superior application availability and increase recovery time in your data center? Server clusters have been proven to preserve availability in the face of failure by creating computing pools capable of rapidly restarting failed VMs. VMware vSphere provides a number of options for clustering in Microsoft-based data centers and supports clustering options for Exchange and SQL Server, making it an obvious alternative for clustering in a Microsoft environment.

Keep in mind that although clustering may seem like a simple solution to preserving HA, it comes with its own unique set of challenges and prerequisites. The key to successful VM clustering is interoperability and frequent cluster architecture evaluation.

How can you make sure an HA server or cluster is working?

Preventing HA server failure is critical to any IT organization, but testing HA clusters can be risky for a number of reasons. Should IT teams choose to test their HA clusters without any sort of backup for their production system, they put themselves at risk of their system going down during testing. Although software exists to protect systems while IT teams run tests, they're often expensive, which makes them unappealing for smaller businesses. So the question is, should IT teams even bother testing their HA clusters or should they put their faith in the system's reliability? Experts seem to agree that testing HA clusters is a necessity, because server and dependency changes can occur at any time, but differ when it comes to the frequency at which tests should be conducted.

Next Steps

Making a high availability data center

Users' guide to high availability

Save money with application-level high availability

Dig Deeper on Virtualized clusters and high-performance computing