The VNX provides network high availability or redundancy by using threetypes of virtual devices that deal with the problem of link or switch failure:
FSNs that extend link failover out into the network by providingswitch-level redundancy.
Ethernet channels that enable multiple active Ethernet connections to the same switch to appear as a single link.
Link aggregation that allows Ethernet ports with similar characteristics to the same switch to be combined into a single virtual device or link.
这个文档有比较详细的描述 https://support.emc.com/docu48505_Configuring_and_Managing_Network_High_Availability_on_VNX_8.1.pdf?...
The VNX provides network high availability or redundancy by using threetypes of virtual devices that deal with the problem of link or switch failure:
FSNs that extend link failover out into the network by providingswitch-level redundancy.
Ethernet channels that enable multiple active Ethernet connections to the same switch to appear as a single link.
Link aggregation that allows Ethernet ports with similar characteristics to the same switch to be combined into a single virtual device or link.
这个文档有比较详细的描述 https://support.emc.com/docu48505_Configuring_and_Managing_Network_High_Availability_on_VNX_8.1.pdf?...
EtherChannel IO的分配机制和负载均衡策略取决于采用的算法,算法规定:如果使用相同的地址和会话信息(MAC,IP相同),则IO总是分布到通道中的同一端口,因此也不是完全的负载均衡。用户可根据需求对算法进行更改,使用基于源/目的MAC地址或IP地址来做负载均衡。链路聚合默认情况下并不真的为各链路速度之和,只是在特定的链路发送特定的报文,给人的感知速度为所有链路的速度总和。
会不会在数据量读取比较大时, 某个IO的传输超过了端口的理论带宽而不会把负载分配给其他端口呢?
不会的,一条链路发生过载时报文不会分散到其他链路,只会丢弃。看了它的实现过程就清楚了,参考这篇文章:链路聚合