Size vMotion bandwidth and estimate live-migration time from VM memory, dirty-page rate and concurrent migration limits.
vMotion copies memory iteratively: a full pass, then repeated passes for pages the guest dirtied, until the remaining delta is small enough to switch over in under a second. The dirty-page multiplier stands in for those extra passes, and because vSphere divides the link across concurrent migrations, raising concurrency shortens the wall clock for a whole host but lengthens each individual migration. Maintenance windows are planned against evacuation time, and a 1 GbE vMotion network running four concurrent migrations gives each one 175 Mbps — under the supported floor, which is how a routine patch night turns into an all-nighter.
vMotion Network
time = VM memory × dirty-page multiplier ÷ (link speed × share ÷ concurrent migrations); the evacuation window is ceil(VMs ÷ concurrent migrations) waves of that time.
time = VM memory × dirty-page multiplier ÷ (link speed × share ÷ concurrent migrations); the evacuation window is ceil(VMs ÷ concurrent migrations) waves of that time. vMotion copies memory iteratively: a full pass, then repeated passes for pages the guest dirtied, until the remaining delta is small enough to switch over in under a second. The dirty-page multiplier stands in for those extra passes, and because vSphere divides the link across concurrent migrations, raising concurrency shortens the wall clock for a whole host but lengthens each individual migration.
Maintenance windows are planned against evacuation time, and a 1 GbE vMotion network running four concurrent migrations gives each one 175 Mbps — under the supported floor, which is how a routine patch night turns into an all-nighter.
This calculator takes 6 inputs: Memory per VM, vMotion network speed, Share of the link available to vMotion, Concurrent migrations, Dirty-page multiplier, VMs to evacuate. The pre-filled defaults are a realistic starting point — replace them with figures from your own environment for a result you can act on.
The cap exists to keep each migration above the 250 Mbps floor. Eight concurrent migrations on a 1 GbE link would leave 125 Mbps each, so the pre-copy phase would never converge on an active VM.