The Ethernet network protocol traditionally has been within the realm of the information technology (IT) department. The IT department may insist that the corporate standard for infrastructure be deployed when the engineering department seeks to place devices that previously were separated from the corporate Ethernet network because they were used on a proprietary fieldbus, onto the corporate network,

Network designs used by IT for business solutions generally have followed a well-established methodology and the network equipment typically has been optimized for the environmental and application requirements of the business. While the IT department understands Ethernet and the corporate network very well, it may not be aware of the unique differences and requirements of Ethernet in an industrial environment.

Industrial Ethernet has several performance and redundancy requirements that set an unusually high benchmark for reliability and recovery in the event of a fault. For example, In a business setting, if network congestion causes a fileserver or email server to delay the completion of a file transfer it typically won't be a problem for end users.

Hundreds of users share the same resources and congestion is not uncommon in a business network. In fact, it may not even be noticed by end users. However, should an industrial Ethernet application experience a similar delay, the result could interrupt the entire application and the resulting production downtime could be very costly to the business.

The Ethernet infrastructure equipment choices play a large role in assuring that the performance and redundancy expectations are met for Industrial Ethernet applications. Acquiescing to the 'corporate standard' for Ethernet switches in an industrial environment may not offer that assurance if the devices are not designed for industrial use.

This article will examine some of the key differences between the infrastructure requirements for traditional business Ethernet systems and industrial Ethernet applications, including differences in architecture, performance, environment, and maintenance.

Network Architecture

Businesses all share similar network architectural requirements. Typical offices, cubicles and meeting rooms drive the port density requirement to the edge of the network. To ensure complete coverage for present and future requirements, overlapping circles are fitted on the floor plan of the building, representing a 90 meter horizontal cabling radius.

Within each circle, closets or rooms are identified to house Ethernet switch stacks, ensuring that no Ethernet cable will exceed the 100 meter requirement. Then, stackable switches are chosen to be located within that radius. As the business grows or changes, stackable switches can accommodate additional ports. The Ethernet closets share the climate control of the building and house standard 19-inch rack frames called Intermediate Distribution Frames (IDFs).

For industrial applications, port density varies widely. Some applications require just a few Ethernet ports in each location and those applications may be several hundred meters apart. Unlike offices, industrial applications have very different node density requirements. Placing a "corporate standard" 24-port Ethernet switch in a location that only requires four ports is expensive and wasteful. Additionally, if the industrial application requires a fiber optic interface that can span three km to the next switch, that limits the available choices for a business Ethernet switch and drives up costs.

Industrial Ethernet switch offers provide the varying port density and interface media requirements that suit the unique architectural requirements for industry. Business Ethernet switch offers are designed for standard, office environment densities. Performance
The most obvious difference between business and industrial Ethernet applications is performance. The business performance requirements are driven by whether there are client"server applications, such as connectivity to a database or a Voice-over-IP (VoIP) phone connecting to VoIP Private Branch Exchange (PBX), versus a core network resource such as an email server. Client-server applications such as database connectivity or VoIP are more critical in terms of connectivity, and this drives the requirement for performance as well as redundancy.

Business client-server applications are more sensitive to packet loss or delays but can still maintain connectivity for several seconds before termination of the session. Applications such as email replication from servers to the desktop client can tolerate more substantial delays without impact. Since resources such as the VoIP PBX, database, email and enterprise file servers typically exist at the network core, redundancy and performance requirements are driven from the edge of the network where the users are, to the core resources.

Traffic flow from the desktops to the servers requires businesses to increase bandwidth from the network edge to the core resources and stipulates the redundancy requirement. For business applications, should a key infrastructure device such as a switch or router experience device or media failure, recovery within seconds is considered acceptable. Redundancy methods employed include the Spanning Tree Protocol (STP), or the evolved derivative, Rapid STP (RSTP).

The timers used in STP or RSTP do not resolve values below one second and therefore can 'heal' the network by finding an alternative route within one to 60 seconds. While this accommodates resources such as email, Internet, Intranet and file services, there is the minor inconvenience of possibly dropping a phone call or having to log back into a database when the session was terminated. The productivity lost is measured in seconds or minutes and only on an individual user basis. Should there be a device or media failure, the impact to the business is localized depending on the activities at the time, and is limited to individual users or resources.

Industrial applications on the other hand, have far more different traffic patterns and sensitivity to recovery time in the event of a switch or media failure. While business applications steer traffic from the edge of the network to the core resources, industrial Ethernet application requirements are peer-to-peer.

Data must be exchanged between peer devices at the network edge so that programmed logic can respond to upstream or downstream changes, as well as share those changes with Supervisory and Data Acquisition (SCADA) systems, Human Machine Interfaces (HMI), Input/Output (I/O) devices and other Programmable Logic Controllers (PLCs). The peer-to-peer nature of industrial Ethernet drives communications and redundancy requirements rather than the edge-to-core requirements used by business applications. Essential requirements
Therefore, the essential requirements for industrial Ethernet applications are:

1) That the recovery time does not affect the production application should a media failure occur

2) That the packet rate of communications be supported by the infrastructure

Regarding the first requirement, media recovery that is over one second is unacceptable in most industrial Ethernet applications. Failed communications for that long of a period stops production which leads to expensive downtime that directly affects the business. Industrial Ethernet switch offers favor a redundant Ethernet ring that provides sub-second recovery time in the event of a switch or device failure. In most cases, this allows the application to continue operating uninterrupted, and allows engineering and maintenance personnel to address the failure while continuing production.

Media or device failure also takes on another dimension in an industrial application. While business systems typically operate on a business hours schedule, many industrial processes are 24/7. A media failure during a third shift on the weekend that allows production to continue alleviates the inconvenience of having to locate and obtain resources to diagnose and repair the failure immediately.

In terms of the second essential requirement, Ethernet switches and devices have evolved to support full duplex communications where any device or switch can transmit and receive simultaneously, and the sheer bandwidth available via Ethernet that is 100 Mbs fast is more than sufficient. In fact, the bandwidth exceeds the capability of the devices to completely consume it.

Additional traffic management services like Quality of Service (QoS) reinforce efficient and reliable message delivery. Broadcast Rate Limiting is also employed to prevent excessive broadcasts from congesting device queues that would delay automation message processing. If a business Ethernet switch is used, features like QoS and Broadcast Rate Limiting may require knowledge of complex switch operating system commands. Industrial Ethernet switch offers on the other hand, have responded by making these features prominent and more easily managed using a Web browser so that engineering or maintenance personnel can perform the configuration.

About the Author

Michael B. Roche, principal technical marketing specialist, Schneider Electric Services, Schneider Electric North American Operating Division