rolffimages - Fotolia
The ability of a virtual environment to utilize multiple physical NICs provides two distinct benefits: load balancing and failover.
Network interface card (NIC) teaming software enables a single virtual switch to access and interact with more than one physical network adapter. Administrators set up NIC teaming policies at the port group level.
When multiple VMs communicate with a single virtual switch, and that virtual switch exchanges traffic with a single NIC, that single NIC port can become a network bandwidth bottleneck that can impact the performance of one or more VMs. NIC teaming software can spread the network traffic from multiple VMs across two or more physical NICs. This creates far greater bandwidth than a single NIC provides.
The most common NIC teaming software option for load balancing is to route traffic based on the switch port ID. Here, traffic from a virtual port is sent to the same physical port unless a failover occurs in the NIC team, and traffic is received by the same physical port.
A route can also be selected based on a source MAC hash, which is a hash of the source MAC address, or on an IP hash, which is a hash of the source and destination IP addresses of each traffic packet. This is often the best option for link aggregation.
A failover configuration enables a virtual switch to send traffic to an alternate physical port when the primary physical port fails. Failover decisions can be made by checking the link status from the network adapter -- such as a failed or removed cable -- but this doesn't detect configuration problems, such as spanning tree errors.
Another approach to failover is beacon probing, which sends and listens for Ethernet broadcast frames, combined with link status, to evaluate link failures and trigger failovers. Normally, NIC teaming software applies a failback policy that will restore traffic to physical ports that become available, putting the backup port in standby once again.
It's worth noting that failover is a powerful tool for network resilience, but isn't enough for a truly resilient network. Best practice dictates a robust network design that builds in multiple redundant pathways for network traffic between critical devices.
Dig Deeper on Network virtualization
Related Q&A from Stephen J. Bigelow
Just because software passes functional tests doesn't mean it works. Dig into stress, load, endurance and other performance tests, and their ... Continue Reading
Don't neglect form factor as part of your data center server selection. Instead, figure out what type of environment you need and learn which server ... Continue Reading
Learn how load balancing in the cloud differs from a traditional network traffic distribution, and explore the different services available from AWS,... Continue Reading