Bridging vs. VIF plugging for VM networking

[ykzeng]

Following the issue #12
We need to rethink VM networking paradigm in EmuEdge, the current VIF direct plug (to an OvS switch) is convenient but limited in several aspects:

1. IP configuration, we cannot gain full control of VM ip address and currently rely on DHCP for ip assignment, with OvS it might be possible to assign fixed ip address to each VM so that the environment is more controlled and more easily to operate
2. Bidirectional traffic shaping: VM-ingress (usually referred to as egress when considering the relative direction to vSwitch) traffic shaping can be done by simply controlling VIF through tc. However, VM-egress(vSwitch ingress) traffic rely on directing traffic to additional ifb and then we controll ifb egress instead to achieve the goal of bidirectional shaping.
3. Performance issues: since we are using an additional ifb that means the path of ingress traffic became longer, which might lead to performance issues to measure.

Using a bridged networking method learnt from the common practice in bridging VMs to external network, might solve the IP configuration issue and bidirectional traffic shaping will be unified, as we can have the VM plugged into its exclusive bridge, then connecting the bridge to other networks through a pair of veth. However, performance might still be influenced, which is left for investigation in the future.

References:
Why Linux doesn't support ingress shaping
ServerFault: hack with ifb for ingress shaping
Bridged Networking of VMWare
How to connect VM network to Physical network interface?