01 February 2005
Redundant Ethernet means redundant Ethernet ports
Regarding December 2004's Networking department speaking to the development of fault-tolerant or redundant industrial Ethernets, fieldbus expert Jonas Berge writes this:
The article defined Ethernet network redundancy as the ability of the network to survive a single cable failure in its switch-to-switch links. Exactly what that definition means is hard to visualize without a drawing.
In the ring topology, a number of local area networks (LAN) switches connect together forming a ring. The Ethernet devices in turn connect to these switches. The LAN switches thus act as a splitter permitting communication to either go clockwise or counter clockwise. Thus, a device having just a single Ethernet port can communicate in either of two directions.
However using ring topology, there are still many single points of failure. For example, a single fault in the device port, the wire between the device and LAN switch, or in one of the LAN switches that connects several devices to the ring would result in the loss of communication for one or more devices. That is, many aspects of ring topology are non-redundant.
Over the past few years, I have found that most users in the process industries do not agree with such a narrow definition of Ethernet network redundancy but take a broader view.
The Ethernet ring topology is not a fully redundant scheme as required in the process industries. Users in the process industries are looking for DCS style redundancy, where the entire network is duplicated leaving no single point of failure.
In other words, redundant Ethernet means redundant Ethernet ports, two cables to each device, and duplicated LAN switches. In such a scheme, the failure can happen on any wire or even a LAN switch without interrupting communications because there are multiple simultaneous paths.
Ring topology is a hardware solution that works independent of protocol and only requires one port per device. It is simple with a minimum of wiring.
On the other hand, full Ethernet redundancy as used in process automation systems requires two ports and two IP addresses per device.
In fact, when using device/card redundancy it is often the case that primary and secondary controllers each have two ports resulting in four ports per device in many cases. Full Ethernet redundancy also requires all the participating devices use the same application layer protocol and LAN redundancy mechanisms.
Most DCSs today use Ethernet and have proprietary application protocols also covering the redundancy. However, the tendency is users also want to integrate some third-party devices from the plethora of Ethernet-based devices and subsystems that have become available.
To enable full LAN redundancy in process control systems integrating third-party devices and various package units, it is necessary to have a standard application layer protocol on top of Ethernet and IP that not only provides a common understanding of data values, but also provides an interoperable LAN redundancy mechanism.
Foundation fieldbus high-speed Ethernet (HSE) is just such a protocol. HSE provides application layer interoperability, not just TCP/IP connectivity, as well as full LAN redundancy, not just ring topology.
Having said all that, Ethernet ring topology is a good idea when the devices only have a single port or if they don't share the same application layer protocol and redundancy mechanism. Ring topology can also mate with real LAN redundancy to form a system with very high availability.
For this reason, users need to pick a standard application layer protocol for Ethernet and IP and apply it throughout the plant. This protocol must support a standard redundancy mechanism with fault detection and routing.
Nicholas Sheble (firstname.lastname@example.org) edits the Networking & Communications department. Jonas Berge (email@example.com) is an engineer with SMAR Singapore Pte Ltd. He is the author of Software for Automation: Architecture, Integration, and Security, ISA Press, 2005 and Fieldbuses for Process Control: Engineering, Operation, and Maintenance, ISA Press, 2001.Return to Previous Page