Keep the vMotion traffic inside the enclosure is a good option, but if you have nodes across enclosures, if you migrate virtual machine between enclosures, the vMotion traffic will leave the enclosure. Anyway, you can control the vMotion traffic with Network I/O or create VMs to Host DRS affinity rules to pin VMs to specific enclosures. And assuming that you have 10Gb uplinks, vMotion traffic will not be an issue higher than keep all nodes inside the same enclosure.