Dispelling myths about industrial Ethernet, part 1

There are numerous myths out there about industrial Ethernet.

There are numerous myths out there about industrial Ethernet.

Ethernet has proven itself as a suitable network for industrial automation and control. Whereas Ethernet as originally codified in the IEEE 802.3 standard is probabilistic instead of deterministic – i.e., as the number of nodes on the network and overall network utilization rise, collisions increase exponentially and access probabilities for given frames decrease at the same rate – the introduction of full-duplex point-to-point Ethernet links like 10Base-FL and 100Base-FX addressed this issue by creating separate send and receive channels and using only two primary nodes for data exchange [1].

Moreover, manufacturers have turned to techniques such as segmentation, via switches and routers, to drive down utilization and in turn mitigate the disruptive effects of Ethernet traffic. Just like its probabilistic character, Ethernet's "noisy" wire – which typically carries many types of traffic, from diagnostics to applications using Web protocols such as HTTP and FTP, in addition to any real-time data – would seem to put it at an inherent disadvantage to serial wires in industrial networking. Segmenting the network via switches, however, is only a start to overcoming this issue, since on a switched LAN local traffic is kept within segment boundaries, even while the segment as a whole retains access to the full performance/data rate [2].

Still, segmentation alone is not a comprehensive solution to keeping Ethernet traffic in check [3]. Switches may not be able to segment malicious, broadcast/multicast and network management traffic and the industrial Ethernet implementation itself must have a star topology as well as managed switches in place. Industrial Ethernet protocols such as Modbus TCP, PROFINET and EtherNet/IP may accordingly use a custom CPU, such as the one at the heart of PriorityChannel, to separate important messages from the rest of the network load based on their ethertypes or designated UDP ports.

In this way, Ethernet's distinctive traits (e.g., its ability to distinguish between different message types) can be turned into decisive advantages for performance on an industrial network. Alongside more apparent perks such as a large vendor ecosystem and straightforward support for open higher level network protocols like TCP/IP, these Ethernet features have made Ethernet-based networks increasingly popular alternatives and supplements to fieldbuses.

Persistence of myths about industrial Ethernet
It is clear, as we have seen, that Ethernet's place in manufacturing plants everywhere is settled. Its market share of industrial networks continues to grow at a faster pace than fieldbus and new developments such as Time Sensitive Networking may further cement its preferred status in automation and control systems. But perhaps because of its roots in non-deterministic networking, there remain many myths about industrial Ethernet floating around.

Let's look at a few of the most common misconceptions about Ethernet in the factory. We'll consider ones about both Ethernet in general and popular protocols such as PROFINET in particular.

Myth #1: Industrial Ethernet is not really deterministic
This myth really comes down to hair-splitting about how collisions are handled and avoided, and getting to the truth of the matter will require separate treatment of Ethernet and the protocols based on it. It is technically true that an Ethernet switch, upon detecting a potential collision, delays one of the packets by a fraction of a second, which may not be truly "deterministic" but nevertheless has no discernible effect on real-time communications.

It would take a massive amount of traffic passing through every single managed switch on an industrial network for the minimal jitter to add up to a significant drag on performance. Such a scenario is made even more unlikely because of the widespread use of the IEEE 1588 Precision Time Protocol, which facilitates submicrosecond clock synchronization over long distances with simple standard cabling. PTP also uses minimal bandwidth and has low overhead, making it ideal for use throughout distributed systems [4].

Once we take up specific protocols, the determinism question is easier to settle. For example, PROFINET can bypass TCP/IP for real-time traffic. Plus, through PROFINET Isochronous Real Time, it can reserve network bandwidth for critical workloads so that those frames are not slowed down even if the overall network load becomes too heavy.

Myth #2: Wireless is not suited for industrial automation and control systems
Wireless options have long been on the fringes of industrial Ethernet networks, being deployed most commonly in situations in which a wired connection is simply not feasible for reaching I/O and controllers positioned on rotating machinery or moving platforms. Also, wireless is often more cost-effective than wired because it does not require any cabling and can be implemented via open standards such as IEEE 802.11 Wi-Fi.

With the emergence of the Industrial Internet of Things, wireless – including Wi-Fi, Bluetooth, cellular and IEEE 802.15.14-based standards such as WirelessHART and ZigBee – is receiving more attention than ever as a way of connecting billions of new devices. But is its performance up to snuff?

Yes. The introduction of 802.11n Wi-Fi in particular was a watershed for wireless in industrial environments. It could match 100 Mbps wired speeds, add signal reflections together to increase range and make full use of Orthogonal Frequency Division Multiplexing to protect against multipath fading in high-speed industrial networks.

Meanwhile, Bluetooth and the 802.15.4 family have also adapted to industrial requirements. Their use of frequency hopping has been especially important for avoiding interference and coexisting peacefully with 802.11 WLANs.

Myth #3: Factory and enterprise IT networks must remain separate for security reasons
Many industrial Ethernet protocols like EtherCAT and EtherNet/IP, to name but two, offer the promise of unifying the network from the factory floor all the way up to enterprise IT. This arrangement can be convenient in terms of cost and operation – industrial Ethernet can replace the disparate standards (i.e., serial connections in industrial applications, standard Ethernet for the office LAN) that would have otherwise coexisted across the plant.

Occasionally, the traditional separation of networks – also known as an air gap – has been justified as a security precaution. It is nowhere near a sufficient measure, though, as demonstrated by the effects of viruses such as Stuxnet and other zero-day exploits targeted at industrial infrastructure. Rather than settle for simple isolation, manufacturers should invest in advanced security solutions and not shy away from unifying the network via Ethernet if appropriate.

[1] http://www.eetimes.com/document.asp?doc_id=1279798

[2] http://www.siemens.fi/pool/products/industry/iadt_is/tuotteet/automaatiotekniikka/teollinen_tiedonsiirto/teollisuus_ethernet/br_ie_switching.pdf

[3] http://www.acromag.com/sites/default/files/White_Paper_What_Is_PriorityChannel.pdf

[4] http://www.ni.com/newsletter/50130/en/

Comments are closed.