Just been given the job of setting up an industrial network?
Selecting appropriate Industrial Ethernet routing hardware can be a daunting task. Robert Wright navigates the options and choices that will largely determine the overall efficiency of the network.
Before investigating all of the features provided by Industrial Ethernet switches, one should first consider exactly what an Ethernet switch is and what it does. An Ethernet switch basically, interconnects Ethernet devices. It receives frames transmitted by one device and passes these frames onto appropriate switch ports, which connect to other Ethernet devices. As it passes these frames it also learns where Ethernet devices are located and uses this information to help decide which ports to use for passing frames. This helps cut down on network use as frames only go to the appropriate switch ports.
Most Ethernet switches use the store and forward method to pass frames. The switch receives an entire frame and then transmits this frame out the appropriate port(s). A hub will receive one bit and retransmit this bit.
As the switch learns where devices are located it stores this address information in a built-in address table memory. It also has memory for frames it is in the process of forwarding. Each switch will vary in address and frame storage space. One consideration relates to the number of Ethernet devices that will exist in an entire network: an appropriate switch should have enough address storage memory for each device in the network. For example if a network will comprise 100 Ethernet devices, the switch should have an address table large enough to store at least 100 Ethernet addresses. Most switches provide basic functionality for monitoring the network in the way of LEDs for link and activity. Most switches also provide the ability to negotiate some of their settings.
Many switches support two types of negotiation. The first is called 'auto-negotiation'. This protocol allows two interconnected Ethernet entities (for example a switch and a computer) to come to some agreement over several parameters used in their communications. One parameter they would negotiate is the data rate to be used. This is generally 10 Mbps or 100 Mbps. They negotiate the use of half or full duplex. Full duplex allows communications to exist in both directions at the same time, while half duplex only allows communications in one direction at a time. They also negotiate the use of flow control in their communications. If flow control is used then either device can request that communications be halted if the device needs time to process received frames.
If supported, Auto-MDIX or Auto-crossover is another negotiation that can occur between two Ethernet entities. This allows these entities to decide which wire pair to use for transmitting frames and which wire pair to use for receiving frames. This feature is attractive when connecting two switches, as this would normally require the use of a special crossover cable. With Auto-MDIX the two switches negotiate which wire pairs to use when communicating and this allows the use of standard (straight through) cables when connecting two switches or two end-devices.
Most Ethernet communications occur over twisted-pair cabling. However, there are times when communications should occur over fibre-optic cabling. These cables are used when signals need to travel over distances greater than the 100 m supported by twisted-pair cabling. Most fibre-optic devices can communicate up to 20 km when using a full-duplex setting. Fibre-optic cables are also used in high noise environments because their communications are unaffected by electrical or magnetic fields. There is a choice of Multi-Mode fibre-optic cable, which will communicate up to 2 km, for longer distances Single-Mode fibre needs to be installed.
Managed versus unmanaged
A key question in choosing an Industrial Ethernet switch is whether to select a managed or an unmanaged switch. A managed switch is basically a switch that supports SNMP (simple network management protocol). Of course most managed switches provide features beyond SNMP but are generally more expensive than the unmanaged variety.
A managed switch allows the user to take control of the network while unmanaged versions simply enable Ethernet devices to communicate: the users connect their Ethernet devices to the unmanaged switch and they usually communicate automatically. There will be status LEDs offering some feedback on link activity but this is generally all there is.
A managed switch will have the same status LEDs but permits the network administrator to adjust the communication parameters to any desired setting and provides monitoring of network behaviour in a number of different ways. For example, in systems that communicate in high noise environments, it is sometimes advantageous to force the data rate to 10 Mbps because noise coupled into the cables may confuse the auto-negotiation process.
Most managed switches permit setting of the data rate for each port. These environments can also benefit from disabling Auto-MDIX support since this negotiation can become confused by noise. Again a managed switch is normally required if you want to enable or disable this feature on a per-port basis.
A managed switch provides viewing of a multitude of network statistics through SNMP. This includes the number of bytes and the number of frames transmitted/received, number of errors and port status viewed on a port basis. Some managed switches also make this data available via a web server so that the user can use a standard browser to view the network status. Advanced features include:
QoS is the ability of the switch to apply a higher priority to certain frames. A switch can use the port on which the frame arrived to determine the frame priority (port QoS) or it can use a tag within the frame to determine its priority (IEEE 802.1p and 802.1Q). These features are useful in improving determinism.
Trunking is two or more ports grouped together and acting as one logical path. This can be used to increase the bandwidth between two switches. Also, in some cases, these paths can provide some redundancy. For example if two 100 Mbps switches are interconnected by two cables then the bandwidth between these two switches can be 200 Mbps.
These allow a switch to logically group devices, and to isolate traffic between these groups even if the devices all share the same physical switch. For example if a switch were being used for both office and factory communications, two VLANs could be created to isolate the office data from the factory traffic. Some switches also allow devices to be located on multiple VLANs - sometimes called overlapping VLANs. This may be used if one device, a scada system for example, needs to communicate to both the office and the factory: this device would be logically present in both the office VLAN and the factory VLAN. The net effect would be to isolate traffic between the remaining office and factory devices but allow the scada system to communicate on both networks.
The traffic isolation advantage of using a switch instead of a hub may become a disadvantage when trying to debug a communications problem. A switch will only send frames to those devices in the conversation. This helps lessen network congestion. However, if the user wants to view all the frames transmitted on a network this feature becomes an issue. Many managed switches offer a feature called port mirroring which allows one port of the switch to monitor the traffic sent/received by one or more ports of the switch. With this feature a PC running a protocol analyser program can capture traffic from one or many ports after port monitoring has been enabled. Protocol analysers are popular problem-solving tools for Ethernet-based networks.
Many switches offer a fault relay to monitor link status of specific ports or the power status of the switch. These dry contacts can then be connected to a PLC or other such control device and used as an input to the control system. This is useful if the user wishes that the control system be alerted when communication to one or more Ethernet devices has failed.
In Ethernet networks there are three types of frames: broadcast frames, which are destined to all devices in the network; directed frames, which are sent to one specific device; and multicast frames, which are sent to one or more devices. Some Ethernet protocols use multicast frames to send data to multiple devices at the same time. These protocols generally create a large amount of multicast traffic. Switches with IGMP snooping can automatically send multicast frames only to those devices that have requested them. This keeps the multicast traffic from flooding unrelated devices. This multicast filtering can be important in large Ethernet/IP networks.
Redundancy is a popular feature in managed Industrial Ethernet switches: if one interconnecting cable were to fail, another cable or set of cables automatically takes over. The time in which this recovery takes place is called the recovery time. There are two popular IEEE redundancy standards, spanning tree protocol (STP) and rapid spanning tree protocol (RSTP). STP (IEEE 802.1D) is the older, slower-to-recover protocol. RSTP (IEEE 802.1w) is a newer and faster version of STP re-establishing connection in 1-2s (vs 30-60s for STP). Because of these long recovery periods, many industrial switch vendors have created their own proprietary ring redundancy protocols. These can generally recover in less than 300 ms. In these networks the switches are connected in a ring. These protocols generally select one switch-to-switch link in the network to be disabled. This is the backup link. When another switch-to-switch link fails the backup link is enabled, thus repairing the network.
Each redundancy method has its pros and cons. STP and RSTP can be wired in a ring or in just about any configuration imaginable. There are several vendors selling STP and RSTP-compatible switches. STP/RSTP networks, however, are generally slower to recover then proprietary ring networks. Proprietary ring protocols must be wired in a ring or in several rings and all switches in the ring must come from the same vendor. Also STP/RSTP networks can provide faster recovery times when used with a mesh network. This requires three connections between switches while the ring network requires only two. Here it is definitely a case where users must talk to their switch vendors about their specific network recovery time requirements.
This is where Industrial Ethernet switches really differentiate themselves from commercial offerings. Industrial Ethernet switches were designed for environments that are not favourable to commercial switches. This can include environments with temperature extremes, high vibration and severe electrical noise. As commercial environments are generally room temperature most commercial switches are designed for a very small temperature range. Also some commercial switches use fans to help in cooling. This could be a problem in many industrial environments due to dust that could accumulate in the fans and the low MTBF of most fans. Commercial switches also expect to be in a low electrical noise environment. This is not the case in many industrial environments. Commercial switches are designed to meet commercial or office EMC requirements while most Industrial Ethernet switches are designed to meet the more stringent industrial EMC requirements.
Environmental concerns can also be as simple as mounting and power supply issues. Most commercial switches are designed for 19" rack mounting or tabletop mounting, but this usually is not acceptable for many industrial environments where DIN-rail mounting or panel mounting is the norm. Also, many commercial switches use a wall-mount power supply. These are generally hard to install in industrial settings and they can be problematic in high-vibration environments where vibrations can dislodge them from standard electrical outlets.
Say, for example, a small system was put together using an unmanaged switch. This may initially be an acceptable situation. However, as more Ethernet devices are added the need for a managed switch may become more apparent. As the user adds Ethernet devices they may find that QoS can be useful to help prioritise the frames due to the increased traffic load of the network. They may also want to use VLANs to help isolate devices an increased number of network nodes. Future Ethernet protocols may well require QoS standards such as 802.1p/802.1Q.
If a move to Ethernet/IP networks is a future possibility, it might make sense to use a switch that supports IGMP snooping. It is possible that the initial system is small and isolated from other networks and does not require the use of IGMP snooping. It may later become desirable to interconnect this system to the company network. A switch with IGMP snooping could help to block this traffic. Determinism Ethernet switches can support a deterministic system that requires knowledge as to how much response and jitter is acceptable in the system.
If I send a frame from one device to another, what is the maximum response time and what is the maximum acceptable jitter?
Most Ethernet switches use store and forward when processing a frame. This means that the entire frame is received and then re-transmitted. If several switches are cascaded, the delay in storing and forwarding these frames increases for every switch. Also, each switch has a small amount of internal latency that adds to this delay.
However, if the system is communicating at 100 Mbps, then the time for the store and forward of each frame is generally on the order of 10 µs for small frames and 130 µs for maximum sized Ethernet frames. Jitter is fairly small - approximately 1 µs per switch - indeed very small compared to the latency of most TCP/IP devices. And protocols such as Ethernet/IP are adding IEEE 1588 (Precision Clock Synchronisation Protocol for Networked Measurement and Control Systems) to help with synchronising Ethernet connected devices.
|Tel:||+27 11 781 0777|
|Articles:||More information and articles about RJ Connect|
© Technews Publishing (Pty) Ltd | All Rights Reserved