During the New England VMware User Group meeting in Newport, RI on April 30, VMware engineers who gave a session called “What’s Next for VMware Virtual Infrastructure” said not to use the upcoming fault tolerance (FT) feature as a general replacement for High Availability because it requires more resources. Instead, only use it where absolutely no downtime can be tolerated.
“To use it, you have to allocate extra resources, and it is a noticeable and significant – you need between 20-30% extra capacity.” VMware has not revealed exactly how much latency results from using its Fault Tolerance feature yet, but engineers said, “You have to expect some latency with the fault tolerance feature, and we continue to work on that.”
Despite its downfalls, fault tolerance is an important feature for IT Pros running mission critical applications on VMs, especially if there are numerous important VMs on a single box, in which case eliminating the risk of losing those VMs during a host failure is critically important. “In most cases the benefits of fault tolerance outweigh the downfalls,” a VMware engineer told attendees.
VMware FT is based on what VMware calls vLockstep technology, which works by having a secondary virtual machine (VM) make an instruction mirror of the first VM on a different physical box, so if a physical component fails, the mirrored VM kicks on and operations continue as if nothing happened.
The two ESX hosts used in FT must share storage systems, and require extra RAM, CPU and I/O capacity, the VMware engineers said.
VMware FT uses VMware HA clusters of up to 16 servers, and any number of VMs in a cluster can be protected with VMware FT.
VMware’s FT feature will be included in vSphere Advanced Edition, which is priced at $2,245 per single processor. The company has yet to announce an official release date.