Industrial Ethernet, enabling seamless integration of office automation and control networks, is one of the most widely used technologies today and is highly popular among users. This article will delve into the development history of Ethernet and its related standards and functions, starting with Ethernet and network protocol (IP). Let's take a look.
Before the advent of Industrial Ethernet, office networks and industrial (factory, process) networks were separate.
The reason is that networks used for manufacturing and processes must be deterministic. Deterministic networks ensure that data packets or messages are always received without exception at specific times. Because Ethernet is a determinate technology, it has no place in industrial network environments. For example, in an Ethernet office network, if an email or file on a server is inaccessible or does not reach its end-user's destination, business continues, and the email or file is eventually saved and accessed when the network becomes available again. On the other hand, if a control signal for a flow rate value is not received at a refinery within a certain time, the tank may overflow, resulting in losses.
The original Ethernet architecture used shared network access and employed an access method called Carrier Sense Multiple Access with Collision Detection (CSMA/CD).
This access method works by "listening" before the terminal node sends a data packet, thus checking if other nodes are transmitting on the network before sending the packet. If the network is busy, the packet will not be sent for a period of time.
If two nodes send data packets at the same time, a collision will occur, and both nodes will need to resend their data packets after a period of time.
In some cases, a broadcast storm can occur when multiple nodes attempt to access the network simultaneously, causing the entire network to freeze.
This shared architecture is also inefficient because every user connected to the Ethernet can see all the requests sent from a single endpoint to "find" the target node. This wastes unnecessary bandwidth, and the situation worsens as the number of nodes in the network increases.
As mentioned above, the initial enabling of Ethernet was highly uncertain and unsuitable for industrial networks, where timing and predictable latency are critical.
In 1990, Kalpana, a company located in Silicon Valley, developed the Ethernet switch. The new switching technology eliminated the previous shared architecture by using an address table based on the end node's Media Access Control (MAC) address.
Media Access Control (MAC) addresses are unique codes that are "burned" into every device that enables the network. The switch builds a table to record which port each MAC address (end device) is located on. After the switch "learns" the port locations of all MAC addresses, it builds source address tables and destination address tables, which are essentially mappings of each user's location.
The resulting point-to-point networks presented new opportunities. For example, when user A wants to communicate with user D, the switch receives a request for a data packet from user A, checks its address table, finds user D on port 3, and forwards the message to user D on port 3. This new technology is more efficient than its shared network predecessor by mitigating network congestion and packet loss caused by collisions.
Switched Ethernet (two collision domains)
The image above shows an S-Switched Ethernet with two independent collision domains. This segmentation is accomplished using the switch's address table. In this table, point-to-point connections eliminate collisions.
With the advent of switched Ethernet, full-duplex communication became possible. Compared to the previous half-duplex Ethernet implementation (where a node could only send or receive data at a time), full-duplex allows a node to send and receive data to and from another connected node simultaneously, thus providing a collision-free environment and doubling the bandwidth capacity.
While switching full-duplex Ethernet enables more deterministic behavior, this new architecture still lacks sufficient resilience or predictability for deterministic industrial networks.
Advancing | Further Development of Ethernet
With the development of switched Ethernet, it was seen that there was a need to further segment the network to improve performance.
The introduction of IEEE 802.1Q Virtual LANs, or VLANs, makes it possible to define logical user groups. These logical user groups can receive data packets from other users in the same VLAN group, meaning that network bandwidth can be utilized more effectively and efficiency can be improved.
VLANs also add a new layer of security because different VLANs cannot communicate with each other without a Layer 3 router, which can be programmed using specific access rules defined by the network administrator.
The diagram above shows a VLAN segment where only users or nodes in VLAN 1 can communicate with each other. This segmentation provides security and traffic optimization.
The Institute of Electrical and Electronics Engineers (IEEE) has released 802.1Q VLANs and 802.1p traffic prioritization schemes, which define up to eight different priority levels (0-7). These priority levels allow more mission-critical traffic to be processed through the network, followed by lower-priority traffic. 802.1p further bridges the gap between deterministic and non-deterministic networks by allowing critical traffic/application networks to access lower-priority traffic.
Another development in Ethernet functionality was the introduction of Internet Group Management Protocol (IGMP) snooping. IGMP snooping is a protocol used by hosts or nodes on an IP network, as well as local multicast routers, to set multicast group membership.
Multicast is a one-to-many architecture; for example, a multicast server sends a single video stream to several different nodes via a multicast router. The advantage of using Internet Group Management Protocol (IGMP) snooping for multicast applications is that only users "registered" for the multicast can receive it, thus freeing up network bandwidth for other services and traffic types.
The Internet Group Management Protocol (IGMP) snooping feature found on many Layer 2 industrial Ethernet switches snoops on the communication between multicast routers and hosts to see which host needs which multicast stream. By knowing which host needs which multicast traffic, the switch can filter out excess multicast traffic from the multicast router, thereby freeing up bandwidth and eliminating network congestion.
To prevent network congestion and improve network performance, Ethernet switches also include additional features such as port-level broadcast, multicast, and unicast storm control. For example, if a network interface card (NIC) on a personal computer malfunctions and starts broadcasting into the network, the switch detects this erroneous behavior and shuts down the switch port to which the personal computer is connected, thus protecting the rest of the network from packet overload and performance degradation.
Rate limiting is another feature that helps maintain network performance, stability, and availability. Rate limiting allows network administrators to set a certain data rate for each port on a switch. This ensures that critical services or devices have more bandwidth on the network than other non-critical services.
When the sending node transmits traffic faster than the receiving switch or node can handle, 802.3x flow control is another method to ensure network stability. In this situation, an overloaded switch sends a pause frame to the sending node, preventing it from transmitting traffic for a specified period until the receiving node has sufficient resources to accept the transmission from the sending node.
Another factor to consider in the development of Ethernet is the increase in speed. Initially, Ethernet operated at speeds of 10 Mbps, then 100 Mbps. Today, many switches offer speeds of 1000 Mbps or 1 Gbps. Furthermore, 10 Mbps interfaces for backbone connections or inter-switch links are also common today. The advantage of these higher data rates is that if errors occur and packets need to be retransmitted, or if network congestion occurs, retransmission is very fast, with virtually no latency. This tremendous increase in speed has made Ethernet a more deterministic architecture suitable for industrial networks.
