Suppose you are tasked with establishing a mission-critical device network that connects your front-line PLCs to your Ethernet-based management system. It's not PLC failure that worries you. What concerns you is the weakest, most vulnerable point in the system - the Ethernet connections themselves. To ensure robust fault tolerance, you need to implement Ethernet network redundancy. The question is, what type of Ethernet redundancy is suitable for your device network, and how should you implement it?
High redundancy, at a price
A common approach to Ethernet redundancy is to establish two physically separate Ethernet networks running parallel to each other. Each device connects to both networks simultaneously. If one network goes down, operations are not disrupted because communication continues through the other network.
This dual-network approach is an easy concept to understand, but the implementation can be tricky. Each device and host must be able to manage the two separate connections and switch over when required, seamlessly and immediately. Advanced protocols and programming are typically required.
Connecting your PLCs to dual Ethernet networks is also a tricky business. Your PLC might not have a single Ethernet port, let alone two. If switching to a specially designed, custom-programmed PLC is not an option, then you'll need to employ a special type of serial port server with two network ports. This kind of serial port server will allow you to connect any RS-485-based PLC, RTU, or other device, to dual Ethernet networks.
Advanced models may offer other features designed especially for redundant Ethernet networks, such as proprietary operation modes, dual MAC addresses, and special PC drivers that automate the negotiation between the two networks
Strong redundancy with lower overhead
Although dual Ethernet networks offer extremely high reliability and can be implemented with single-host or dual-host architectures, it requires high overhead, double the amount of Ethernet wiring, and a considerable investment in software and application development. Another approach to Ethernet redundancy is to connect devices in a ring architecture. This single-network approach offers surprisingly robust fault tolerance. Any single connection can fail without disrupting communication between all devices.
It's a lot easier to connect PLCs to an Ethernet ring because only a single connection is needed. PLCs with an Ethernet port or module can connect to an Ethernet switch on the ring.
If your PLC is RS-485-based, you can use a standard serial port server to connect to the Ethernet switch.
Even better, you may be able to forego the Ethernet switch altogether and use a specially designed serial port server with built-in support for redundant Ethernet rings.
The beauty of this approach is greatly reduced overhead, cost, complexity, space requirements, and time to implementation. Although there is no device or host redundancy in this type of approach, for many situations, this is an acceptable and attractive compromise.
Let's see how these approaches can be applied to real-world scenarios. At a power substation, suppose you have PC-based management stations monitoring a set of RTUs. In this case, you've determined that you need both PC redundancy and Ethernet network redundancy. This means you'll be implementing a dual host dual network system.
Your RTU will connect to a specially-designed serial port server with two Ethernet ports. The serial port server will connect to each of the two networks and will require additional configuration and/or programming to negotiate network switchovers. A single failure in any PC, Ethernet cable, or Ethernet switch will not disrupt the system.
On a train, suppose you are installing simple data acquisition controllers at each car that monitor temperature and other environmental metrics. For enhanced reliability with a minimum of fuss, you can employ serial port servers with built-in Ethernet ring functions. Each data acquisition controller connects to a serial port server, and the serial port servers connect to each other in an Ethernet ring topology. The entire system remains operational even when there is a complete failure in one of the Ethernet cables or connections.
There are a number of effective ways to achieve better fault tolerance for your device network. While complete device and network redundancy provides the most robust fault tolerance for a system, it also involves considerable complexity as well as high implementation costs. A simpler alternative can be to use a ring topology to provide strong fault tolerance for the network connections to your PLCs or RTUs. Since it is the network connections themselves that tend to be the most vulnerable, this approach can be very effective in maximizing overall system reliability with less effort. The approach that suits your application will of course depend on your particular requirements.
» Back to index