- Power cost of managing devices
- Predictable and deterministic routing decisions in the case of failover
- Optimal routing decisions based on bandwidth and load
The way this is done will depend on where in the network the NFV functionality is hosted. There are three basic methods:
- Standard servers running NFV functionality in the same topological network location as the appliance. This saves on hardware costs but does not reduce network complexity.
- Much of the network functionality, including switches and appliances, can be virtualized in a hypervisor. This is the approach taken by Nicira and others. This may be appropriate in data center environments.
- OpenFlow-capable switches can selectively shunt traffic to standard servers running NFV functionality off the main data path.
沒有留言:
張貼留言