Introduction To A Monthly Section “Industrial Networking”
Industrial Networking is becoming an increasingly important element in manufacturing. Manufacturers’ Mart has collaborated with a local firm (ICT) focusing on Industrial Networking to review areas of this technology. Each issue will present topics that will help readers gain a better understanding of networking from Device level applications, Fieldbus technologies, and the emergence of Industrial Ethernet. Readers are encouraged to submit questions or bring out issues to markc@ictglobal.com. Each issue will try to address some aspect of this important topic.

Determinism Is Often Misunderstood
A couple of years ago, when Ethernet first was introduced to the Industrial Automation market, one could not engage in a conversation with anyone in the industry without the "D" word monopolizing the conversation. That "D" word, as we all know, is “determinism”.

When making my first plunge into this "new to me" market, I was told that Ethernet wouldn't work with factory types of applications, as it was non-deterministic. In several presentations to audiences, this recurring theme raised its ugly head. Then, I asked two people from different industries to give me their definitions of determinism. The result was two very inconclusive definitions of determinism and its causes although they both knew it was an issue. It was very apparent that this was the position of the vast majority of the Industrial Automation sector. It was an issue but no one was ‘really’ sure why.

Determinism is defined as " the ability to send a piece of information to a destination and receive a response in a repeatable time frame". One can substitute a plethora of synonyms and substitute words here and there but the crux of the definition remains.

People were confused because they were accustomed to the original Ethernet. You know, the one drawn on the napkin of the PARC in the early 70's.



When the technology was developed, there were no client/server models let alone any network intensive applications, no high-speed video applications (I think those were called VCR's back then) and certainly no VoIP (voice over ethernet ip) technology. We were using the attributes of a technology developed 30 years ago and thinking about it in terms of the whirlwind technologies of the next millennium! It was then that I changed my methodology of presenting and confronting the "D" word issue.

I began by teaching the original Ethernet and the associated hardware and progressing through the client/server world to the evolving technologies and showing the technical and hardware changes along the way, ending with the technologies that we presently utilize.

It was then that many clients began to understand the issues surrounding the "D" word. Slowly, their fears were replaced by uncertainty and the uncertainty by a clearer understanding.

Here, I will cover the most pertinent topic - shared vs switched networks and how the data flows from source to destination. Once this is understood, it is easy to understand the fuzzy logic surrounding the "D" word.


Shared Networks vs. Switched Networks
When Ethernet was first developed, it was with the mindset of "sharing" information in the same geography. It was, very simply, a common medium (thick-wire Ethernet) with devices connected directly to it with a MAU (Media Attachment Unit) It shared a channel of a given frequency (10Mb). It is a source/destination-based arbitrary method of access to the network. This means that every device has a unique address and, as long as it is valid, the device can access the network anytime the channel is free. (See figure 1 – 1)

The term "shared" had more than one meaning, however. Not only did all devices share information and bandwidth, they also shared the messaging. The devices directly connected to the network have no way of knowing "where" the destination node is located. The only way to ensure the information gets there is to send the info to every connected node on the network. (See figure 1 – 2) Thus, all share the excess traffic and network overhead along with the messaging. Hence, more devices on the cable led to bigger networks and to more overhead and problems.

This results in a longer packet delivery time as there are more nodes contending for bandwidth, trying to push messages through a busy network. Ever notice that when you try to log in to the network early in the day when there are few users it seems ‘really fast’? Then, after lunch, when everyone is checking messages, the network is ‘really slow’? The reason is that there is less available bandwidth and more overhead. Simple. Shared Ethernet has variable delays (non deterministic) as opposed to fixed delays (deterministic). As more devices and overhead are added to the network the delay time becomes more unpredictable.

Switched Networks
As the networks grew and technology improved, methods were developed to regulate traffic flow and reduce overhead (Bridging). Bridges were traditionally two-port devices that connected LAN A to LAN B. (See figure 1 – 3) This helped by keeping traffic from one network off a network on which it did not belong. However, networks were growing exponentially and there were still problems. Applications and server-based technologies were constantly being developed and were very network bandwidth-intensive, utilizing lots of available resources. At around the same time, faster speeds of Ethernet were being deployed.

The Ethernet switch replaced the traditional two-port bridge. The technology remained basically the same with the addition of multiple ports. In this way, virtually every connected device could be given dedicated bandwidth. In essence, each device is on its own "network". Like bridging, when a user wants to send a piece of information to another device, it is directed to that device and only that device (actually to the port to which the device is connected).

Many industrial applications are time-sensitive and require a response time in milliseconds. It is recommended by industry that critical, time-sensitive devices be kept dedicated to a switched port in order to minimize the possibility of collisions and utilize the maximum amount of usable bandwidth thus allowing more reliable data delivery.
Even from this brief description, one can see how naturally overhead can be reduced, how much bandwidth is increased and how networks can be segmented and micro segmented to create more throughput and usable bandwidth, increasing the effectiveness of network resources.




About the Author
Mark Cotter, Sr. Systems Engineer, Industrial Communication Technologies Inc., has twenty years experience in networking and communications with the past four years focused on Industrial Automation. He has worked for Hirschmann Electronics, a major manufacturer of Automation and Networking equipment, as a Technical Sales Manager and Engineer.

He has developed and performed many training courses for clients and spoken at major trade shows and written a series of articles on Ethernet communications. He is also a Certified Cisco Networking Professional.