A little history: EtherCAT was initially developed by Beckhoff Automation and first introduced in early 2003. It was intended to use Ethernet for automation applications that required relatively short cycle times of less than 100 microseconds.
How it works: An EtherCAT master sends out a message that passes through each node on the network. Then each EtherCAT device reads the data in the message that is addressed to it specifically, doing so “on the fly.” It then inserts its data into the frame as it moves downstream. So in effect the frame is only delayed by hardware propagation delay times. And because the EtherCAT master is the only node allowed to actively send an EtherCAT frame (with other nodes only allowed to forward frames), this prevents unpredictable delays, guaranteeing real-time performance. It also eliminates protocol stacks such as TCP/IP, which serves to increase transmission speeds and making it more real-time in function.
In terms of real numbers, an EtherCAT network can process roughly 1,000 I/Os in about 30 µsec, or approximately 100 axes in 125 µsec.