Abstract
Keywords
Introduction
Ethernet and industrial Ethernet are increasingly being applied in the field of industrial control.1–4 Various approaches from hardware and software, such as hardware updating, improvement of protocol, optimization of scheduling strategies, and so on, have been tried and used to further improve efficiency of industrial network. Prominent advance has been made in this way. Now latest research has shown that flexible management of network architecture can further improve network performance in a quite different way from those mentioned above.5–7 Recently, new fashions for manufacturing, such as industry 4.0, have been proposed by some European nations that are advanced in manufacturing industry to cope with challenges from limited resources and environment around the world. Fashion for manufacturing is changing from concentrated to distributed to make manufacturing more flexible and digital. A new industrial control architecture is expected to be built. Development of modern network technology provided the opportunities for such industrial control architecture. Application of network technology in industrial control transformed the control architecture from conventional simple closed-loop control to large-scale networked control based on fieldbus, industrial Ethernet, and even wireless sensor network (WSN). New industrial control network is expected to meet demands such as high efficiency, high intelligence, scalability, and re-configurability. Software-Defined Network (SDN) can meet such demands.
Currently used industrial control network
Generally, layered architecture is widely adopted in current industrial control network. There are four layers including field layer, control layer, supervision layer, and management layer in the architecture. The architecture is shown in Figure 1. For convenience, we just name this kind of architecture as CICN (currently used industrial control network).
Currently used industrial control architecture.
Ethernet TCP/IP is used in management layer, the toppest layer in the architecture. This layer is used to manage the whole control process and massive data. Data mining and decision-making are supported in this layer. It can also serve as the bridge connecting upstream and downstream in a whole industrial production chain. Commonly used equipment in this layer are file server, application server, data server, and access to the Internet. No time critical data transmission and operation are required in this layer, and part of the data is commerce related. Its corresponding network can be characterized as Soft Real Time (SRT).
Communication in control and supervision layer is based on industrial Ethernet, and WSN is also used in these layers sometimes. Hybrid communication fashion consisting of wired and wireless modes is used in these layers to connect the field and control center. These serve as the fundamental equipment in industrial process automation. Ethernet is used as the backbone network and the WSN mainly consists of long distance multi-hop mesh. Time critical data transmission and operation are required in this layer, such as cyclic data transmission from field to the operator workstations (OWS) for supervision. Some non-time critical traffic services, such as some diagnostic data from the process control units to the OWS for analysis and documentation, are also provided in this layer. Its corresponding network can be characterized as Weak Hard Real Time (WHRT).
The lowest is the field layer, which is directly connected to the field. Various sensors, transmitters, and intelligent instruments are included in this layer. These devices are interconnected with industrial control systems via various fieldbus, such as Foundation Fieldbus and Profibus conforming to industrial data transmission protocols. The main function of this layer is to connect various field-based sensors and controllers such as intelligent instruments, switches, and valves to the backbone network. The field communications for supervision purpose and control purpose are both adopted in this layer. The field communication for supervision is via low-speed fieldbus such as FF-H1, while the field communication for control is via high-speed fieldbus such as FF High-Speed Ethernet (HSE) and Profinet.
SDN with its latest development
Introduction of SDN
Traditionally, control and forward of network traffic are wholly dependent on network equipment where operating system and proprietary hardware from different manufacturers are integrated, which meant that the control of the whole network was quite difficult or even impossible. Architecture of network was the key point in the success of Internet. However, with the increasing enlargement in the scale of network, too many complex network protocols were installed in closed network devices to make it much more difficult for Internet service providers (ISP) to optimize network and impossible for technicians to deploy new network protocols. Conventional network devices, such as switches and routers, were controlled and locked by their respective manufacturers. Thus, it was hoped that network control and physical network topology can be separated in SDN to remove the limitation to network architecture from the hardware. Consequently, users and enterprises can modify network architecture just like updating and installing software to meet the requirements in adjusting, enlarging, and updating network architecture while keeping the switch or routers in the basic level unchanged to save cost and time in network implementation. SDN is a novel network architecture and first proposed by Emulex. Flexible control for network traffic can be achieved with SDN by separating device control from equipment data, which provided a good platform for core network with its application. Bottom-layer hardware of a specified network is controlled by software platform in the concentrated controller in programmable fashion in SDN to implement flexible configuration of network resources on demand. Comparison of traditional network architecture and that of SDN is shown in Figure 2.
Traditional network architecture and SDN network architecture.
The following five elements are included in SDN architecture:
Controller—All the equipment in the whole network are managed by controller. The whole network is virtualized as a resource spool, and controller can allocate resources flexibly and dynamically on demand of different users and based on global network topology. Controller communicates with the fundamental network layer with standard protocol and provides application layer with control capability with open interfaces.
Physical layer—Physical layer is represented with hardware equipment layer and purely focused on transmission and forwarding of data and transaction. Safe and reliable communication with controller layer is mainly concerned in this layer. The processing ability in this layer is necessarily high to achieve high-speed transmission and forwarding of data.
Southern interface—Southern interface serves as the channel for data transmission between physical equipment and controller. Related data representing equipment status, data flow chart, and control instructions are all conveyed via southern interface to control and manage equipment.
Northern interface—Northern interface serves as an open interface to upper-level transaction via controller, aiming to enable applications to conveniently call the resources from bottom layer. It directly serves application layer. Close connection to transactions is required in the design of the northern interface. It has the characteristics of variety.
SDN application layer—SDN application layer programs the bottom-layer equipment with the interface provided by controller layer.
The architecture of SDN involving the five elements is shown in Figure 3.
Architecture of SDN.
Network abstraction layer and network virtual layer were included in SDN. Concentrated control was implemented in network abstraction layer and resource configuration was implemented in network virtual layer. The kernel technology of SDN was OpenFlow. OpenFlow-based switches in SDN took the place of switches and routers in traditional network. The comparison of traditional network and SDN is shown in Table 1.
Comparison of traditional network and SDN.
OpenFlow was an open and mainstream specification to regulate the data communication in SDN. Four parts were included in OpenFlow specification, which were OpenFlow ports, OpenFlow flow table, traffic channel, and OpenFlow protocol and related data structure. Figure 4 illustrates the OpenFlow mechanism.
OpenFlow mechanism. 3
There were three layers in a typical SDN structure. The topmost layer was application layer including various different transactions and applications. The medium control layer was mainly responsible for the maintenance of network topology and status information. The bottommost layer was responsible for the data handling, forwarding, and acquisition based on flow table. The flow table can be viewed as an abstraction of the data forwarding function by network devices. As for conventional network equipment, data forwarding by the switches and routers depends on MAC address table as well as IP address routing table. It is also the case for switches based on flow table except that the configuration for the whole network is also included in flow table, which makes a richer rule set for data forwarding. A flow table in SDN controller can be viewed as the enhancement and improvement of a routing table in conventional network equipment. Essentially, SDN possessed three characteristics, which were separation of control and forwarding, virtualization of device resources, and general programming for hardware and software, bringing merits as follows:
Generalization of hardware device. Hardware only focused on data forwarding and storage and has nothing to do with service characteristic. The hardware architecture can be implemented with commercial choice.
Intelligence of network can be realized completely with software. The categories and functions of network equipment depend on configuration by software. The operation and control of network can be implemented with network operating system (NOS).
The response to service is much faster. Various network parameters, such as route, scheme, QoS, and stream control, can be customized and configured into the network, which can significantly reduce the time delay before service is opened.
Latest development of SDN
SDN has been proposed and developed for several years. SDN enabled the separation of control from switching hardware. It separated network control from the data transmitting hardware that used to be packaged in the same equipment, with the goal to make the network programmable and to enable applications and network services to directly control the abstracted infrastructure. OpenFlow was introduced as an important protocol, enabling the necessary communication between the control layer and the network elements in the infrastructure layer.4–6
An update and extension of SDN was published early in 2016, providing better understanding and more insight into the tasks and functionality of an SDN controller, as well as the relationship between the SDN controller and the rest of the networking environment. This resulting SDN architecture is applicable to all kinds of applications across the enterprise, carrier, data center, and campus network areas, from end customer to hardware owner, for both completely new and evolving existing network deployments. It accommodates SDN within and between different domains. 8 The major components of the latest version of SDN are resources and controllers. It enables the flexible support of a broad range of use cases and scenarios on a common infrastructure, tailoring and optimizing services according to their diverse requirements. The latest version merges two models based on distinct perspectives. Figures 5 and 6 illustrate the role of SDN controller in the new version.7–13
SDN controller as a feedback node. 2
Controller at the core of the SDN architecture. 2
The controller is the core components of the SDN. It is easy for the controller to be a bottle neck for the whole SDN system with the scale of SDN increased. Therefore, demands for the reliability of controller is becoming higher and higher. Hardware and software redundancy should be considered and implemented for the controller to avoid the lowered SDN reliability due to the invalidity of controller.
Architecture of industrial control network based on SDN
CICN has been used for many years in industry. However, it is becoming more and more difficult to meet the requirements of increasingly developed industry. The disadvantages of CICN have being exposed gradually. Typically, distributed control and management fashion are used in CICN. Traditional query mode for MAC address table and routing table is widely adopted in CICN. The communication protocols from various manufacturers of industrial products are quite different and most of them are not compliant with each other. The control scheme and management mechanism of network are predefined and pre-configured. The communication over the network has to be cut-off to re-configure and re-define parameters of network, if new demands are rising. And, procedures link editing, compilation, and starting have to be redone for the concerned hardware and software of the network.
SDN-based industrial control network can solve the problem properly. For convenience, we just name the SDN-based industrial control network as SDNICN, which is shown in Figure 7.
SDN-based industrial control network (SDNICN).
Typically, intelligent network components, such as industrial switch, wireless industrial switch, industrial network controller (NC), service and management center, and service cloud platform, are included in an SDNICN. Switches for industrial purpose play important role in SDNICN by providing fundamental network interconnection for the whole industrial control network. As for the interconnection between service cloud platform in the concerned industrial network, NC is used for data transmission, forwarding and routing control, and management between different layers. Service Management Center (SMC) is essentially responsible for managing various services used in industrial process control. These services are equipment-oriented and corresponding to various modules of software for control and communication. Devices and equipment in the lower layer of industrial process architecture, such as intelligent instruments, Programmable Logic Controller (PLC), and industrial robots, can be controlled and managed by SMC. Those functions implemented in corresponding industrial process by those devices and equipment can be wrapped into various library modules, which can be called by other component in the whole system. 10 The cloud service platform consisted of distributed control system, manufacturing execution system, enterprise resource project, and other related industrial systems. The cloud service is customer service software that delivers in the cloud and can be called anywhere over Internet. With the cloud service, almost 100% of industrial customer cases in real time from anywhere can be handled. The main functions of the cloud service platform in this architecture include data requisition, resource management, task management, equipment maintenance, and fault diagnosis. The implementation procedures of these functions were designed based on Quality of Control (QoC) and Quality of Service (QoS) of the whole system. Just like those service modules in SMC, the wrapped library modules to be called to implement process production can also be provided by the concerned control systems and management systems in the cloud service platform.
The NC in this architecture can be viewed as the channel connecting control equipment, such as DCS and MES, and the industrial field in the whole architecture. Control action can be initialized by components in cloud service platform. SMC can require various parameters, variables, status, and data needed from the industrial control network via the north interface of the NC. Service modules of the field equipment or devices can be called by SMC via the north interface of the NC aiming to actuate those field equipment and devices. Thus, a complete operational procedure can be formed.
The NC defines the data flows that occur in the SDN data plane. Each flow through the network must first get permission from the controller, which verifies that the communication is permissible by the network policy. After acquiring the corresponding operational procedure for data transmission and application, NC can configure and set related tables or lists to related industrial switches via the south interface according to requirements of the operational procedure, to implement control and management of the whole network. If NC allows a flow, it computes a route for the flow to take and adds an entry for that flow in each of the switches along the path. That means that the NC manages the forwarding state of the switches in the SDN. This management is done through a vendor-neutral Application Programming Interface (API) that allows NC to address a wide variety of operator requirements without changing any of the lower-level aspects of the industrial network, including topology. With all complex functions subsumed by the controller, switches simply manage flow tables whose entries can be populated only by the controller. Based on users’ demands and industrial production plan, network administrator can customize corresponding forwarding and control fashions for the network control to meet different demands. Tasks dealing with network that needed to be processed in switches and terminals in traditional industrial control, such as routing, forwarding, and traffic control, can all be implemented in NC by means of software definition. The network configurations by software definition can also be modified as required in any time. The network configurations by software definition are represented by flow table in industrial switches and can be adjusted and configured according to requirements from production plan. OpenFlow specification is used in the communication between the controller and the switches.
NC can provide a highly extensible and high-performance control plane in the whole architecture. User-friendly human–machine interface (HMI) in Web style can also be obtained via graphical user interface (GUI) fashion in NC for monitoring of industrial process. NC with its north interface decoupled control and network management, while the south interface isolated NC from field devices. All these provide users with more flexibility, convenience, and reliability.
The SDN-based industrial control architecture is remarkably flexible; it can operate with different types of switches and at different protocol layers. NC and switches can be implemented for Ethernet switches (Layer 2 in ISO/OSI model), Internet routers (Layer 3 in ISO/OSI model), transport (Layer 4 in ISO/OSI model) switching, or application-layer switching and routing. SDN relies on the common functions found on networking devices, which essentially involve forwarding packets based on some form of flow definition. With the decoupling of the control and data planes, SDN-based industrial control architecture enables applications to deal with a single abstracted network device without concern for the details of how the device operates. Network applications interact with NC in a simplest way like API. Thus, it is possible to quickly create and deploy new applications to orchestrate network traffic flow to meet specific control requirements for performance or security. 14
Last but not the least, interfacing of field intelligent devices to the SDN-based industrial control network has to be noted. Today, most of intelligent sensors and actuators can be connected to Ethernet via TCP/IP protocol, which makes the system integration easy to be implemented. However, there also some of field devices are still working via protocols different from Ethernet, such as RS232/422/485 or even analog signal. As for these cases, a protocol converter or gateway can be installed between to connect these devices to Ethernet-based SDN industrial network, as illustrated in Figure 8.
Interfacing of field intelligent devices to the SDN-based industrial control network.
Conclusion
SDN-based industrial control network prominently improved real-time performance, re-configurability, flexibility, and compliance of the CICN, meeting demands of intelligence and informatization for control network from industry. Difficult and time-consuming tasks in industrial networked control, such as flexible adjustment of network architecture, security mechanism, and traffic management fashion, can be solved essentially, which means great significance for the future of modern industry.