From the vSphere perspective, your isloation will be accomplished using Port Groups. You would create one or more port groups for your VM's, one for vMotion, two for iSCSI (for multipathing), one for management traffic, and so on. In your vSwitch or dvSwitch settings (depending on whether you're using standard or distrbuted switches), you can do crude traffic shaping on a portgroup level and set peak/burst settings, though this probably isn't recommended.
If you are licensed for Enterprise Plus on version 5.x and are using distributed switches, you can use Network I/O control and use network resource pools to manage allocating available bandwidth per application - this would be the VMware recommended method.
Much like with CPU and Memory, it is not recommended to set hard reservations or limits on network resources unless you have a good reason...