What is static link aggregation?
Static link aggregation is a method of combining or bundling of multiple switch ports or nics to form a single etherchannel. It increases throughput and redundancy for the two connecting devices. A static link aggregation means that you manually configure the links you wish to use as part of your etherchannel.
What is LACP?
Many think of LACP as link aggregation, but it’s actually a protocol that runs on top of the link aggregation. LACP stands for Link Aggregation Control Protocol which allows a network device to negotiate an automatic bundling of links by sending LACPDUS. LACP can be thought of as a protocol that is running over a link aggregation group but it is not necessarily required.
What are the similarities?
- On a Cisco switch setting up LACP vs static port channel is a difference of one word in the following command:
channel-group 1 mode on
channel-group 1 mode active
- You can use load balancing on a static link aggregation group or a LACP link aggregation group. They are independent of each other and can be configured differently on either ends
So why use LACP?
- Automatic failover – When using media converters in a static link aggregation, a failure of a single link will not be detected by the peer on the other side. The other peer will continue to send packets without noticing. Since LACP sends LACPDUS the other peer will be informed to disable the corresponding link. This problem can also occur when using fiber cables as a medium. Since fiber cables have independent lanes for transmit and receive, a damaged cable can cause a unidirectional link to be formed. This can go undetected without features such as UDLD which requires a Cisco device on both ends. With LACP the error can be detected much sooner.
- Mitigates configuration errors – Configuration errors can go undetected in a static etherchannel where as with LACP it is much easier to troubleshoot the any configuration errors
- Prevents premature links to be formed – Links formed using a static link aggregation can be brought up at various times causing traffic to be sent out prematurely. LACP prevents this from happening as it confirms that both peers are ready before traffic is being sent.
How do you setup LACP?
Please visit my previous post on how to setup etherchannel with LACP on Linux and Cisco devices.
The importance of using LACP on VMWare
VMWare didn’t introduce LACP until version 5.1 onwards, and little admins know the importance of using this feature when combined with route based on IP hashed. As I had previously explained, when static etherchannel is used on the switch, traffic will flow down all ports as soon as the link status is up. This means that if you had misconfigured one of your ports on your VM host to not participate in the port group you could end up with a unidirectional link rendering a switching blackhole. If this non participating link happens to be up, but the switch is unaware that it’s not participating in the port group, the switch will continue to send packets down the path even though the virtual switch is not going to process it.
If you had LACP configured, this would have never happened as the link that does not speak LACP would be in a suspended state along with the uplink switch port.