VMware provides a range of availability features to protect virtual machines: HA, FT, VADP, SRM and vMotion. The...
key to maximizing virtual system availability lies in knowing which features to use based on company policies and available technologies in your infrastructure. Here's an overview of what to use when.
Unexpected host downtime: VMware HA vs. FT
VMware vSphere High Availability (HA) is by far the company's easiest availability technology to implement. If you simply have shared storage and two or more hosts configured in a vCenter cluster, you can enable HA. VMware HA will reserve enough capacity to tolerate the loss of one or more hosts and automatically restart virtual machines (VMs) that were running on a failed host. This feature will quickly recover VMs with a minimal amount of downtime.
If you choose VMware vSphere Fault Tolerance (FT), it can eliminate the small amount of downtime that HA requires. When you enable FT on a VM, a shadow version of the VM is created on a second host. As the primary VM executes transactions, the secondary version executes the exact same transactions. The virtual machines are exact duplicates, except that vSphere blocks any writes to disk or network-based communications from the shadow copy. If the host containing the primary VM fails, the second host enables full write access and network connectivity to the secondary VM. This transition is quick enough that the VM's applications are unaffected.
There are a few caveats to using VMware FT, most notably that any protected virtual machines can only be configured with a single vCPU and that each host can only handle four protected VMs.
Expected host downtime: VMware vMotion
When a VMware administrator takes a host offline, vMotion can be used to pass VM execution from one host to another. Any downtime during a vMotion operation is usually the loss of just a few network packets, which should be well within the tolerance of any TCP/IP-based application. VMware vMotion is an indispensable feature for maintaining critical applications' uptime.
VM backup and recovery: VMware VADP vs. SRM
VMware introduced vStorage APIs for Data Protection (VADP) to give vSphere users a standardized application programming interface (API) set for backup applications to directly access VM files. If a backup tool vendor chooses to take advantage of it, VADP's Changed Block Tracking (CBT) feature enables the host to track which blocks have changed within the VM files since the last backup. CBT can significantly reduce the time spent on an incremental VM backup. Though it does not automate VM recovery, VADP allows an organization to back up VMs into a format that is easy to restore and execute with minimal configuration changes.
To automate recovery for an entire vSphere-based data center, check out VMware vCenter Site Recovery Manager (SRM), an add-on VMware product. SRM depends on storage replication, either vSphere or array-based, to duplicate VMs at the recovery site. You'll need a full vSphere infrastructure at the recovery site for SRM. Once the data is available at the recovery site, SRM automates all the steps to make storage available to the recovery site's vSphere hosts, register and reconfigure the VMs and power them on in the appropriate sequence. SRM can drastically reduce the downtime associated with a data center failure, compared with other backup and recovery plans. Be sure to invest time on the business side of SRM plans, not just the technological aspect.
Using all these VMware availability features in the proper scenarios will drastically improve the availability of virtual machine-based applications.