High-speed memory sharing improves application reliability
Control Engineering International: Data transmission based on the Reflective Memory from GE Intelligent Platforms provides a unique industrial real-time network technology that can boost reliability for many industrial control applications.
In the industrial automation application area of cold rolling and hot rolling, the requirement for pressure and rotating speed is strict and demanding. Data must be transmitted to the next node within a very short time to achieve real-time parameter coordination. A very short delay could cause errors and huge waste. Currently, fieldbus and Ethernet cannot satisfy the requirements of high determinacy and timing. Reflective Memory from GE Intelligent Platforms targets demanding simulation, process control, and data collection to make up for this insufficiency and fulfills strict real-time requirements with determinacy, low latency, and high-speed memory sharing.
Reflective memory is a high-speed network with a 2.12 G transmission rate; its transmission speed can be as high as 174 Mbytes/s. Shao Jianfeng, an embedded system application engineer for GE Intelligent Platforms, said that in the reflective memory optical fiber ring, high-speed synchronization would transmit data to the next node in the network and get ready to insert data at any node any time data was written to a local reflective memory device. Each node receives data from its previous node, decodes the data packet, checks the errors, writes this new data to local backup, and sends it to its next node. When data returns to the starting node, it is deleted from the network. In this way, every computer holds the latest local backup of the share memory collection with no software latency and negligible hardware latency. As the example diagram shows, all computers can receive the data written to reflective memory within only 2.1μs. This low latency is vital for building real-time systems (such as simulators, PLC controllers, testing platforms, and high-availability systems).
In the system, all CPUs writing to shared memory will be duplicated to all other nodes, up to 256 computers within the network. All subsystems have sufficient and unlimited access authorities.
Beside a ring structure, star topology of reflective memory network is another option, which enables higher synchronization. Optical fiber hub can bypass any nodes that terminate the operation even if the node interruption is turned off. Thus every computer always holds the latest backup of share memory collection within the network. Backup node can seamlessly take up the operation using the failed node, reducing the negative impact on the productivity, profitability, and performance caused by unexpected shutdown.
Can it be replaced?
GE Reflective Memory provides a ring-structured network for data insertion using optical fiber with 2.12 G transmission rate. The range between nodes can be up to 10 km (single mode)/300 meters (multiple mode).
“Compared with Gigabit Ethernet, reflective memory has a higher real-time performance. The latency between two nodes is no more than 750 nanoseconds, while Ethernet and fieldbus cannot achieve this, as the speed of Gigabit Ethernet (including UDP) is only 100 MB/s,” Jianfeng said. (See comparison table.) It is difficult to achieve similar latency using Ethernet as well as other network technologies, due to constraints, such as IP protocol cost, addressing, and time to write to memory.
Then, can 10 Gigabit Ethernet replace a Reflective Memory network? Jianfeng said that most industrial fieldbuses are still using 10/100 Mbit/s, and currently Gigabit network and 10 Gigabit application are mainly used at server levels. It is possible that a 10 Gigabit network will be used widely in industrial field applications, but that will take a long time. Meanwhile, GE also is actively developing 10 G optical fiber to fulfill environmental needs for more demanding real-time performance.
On the other hand, Reflective Memory has another advantage as it doesn’t rely on any network protocol, which ensures connectivity without additional load limit or terminal rules.
“Reflective memory hardware can be used in VME (versa module eurocard), PCI/PCI-X, PMC (PCI mezzanine card), PCI Express, and others, which allow the separate reflective memory network to connect to a different bus. [PCI stands for peripheral component interface.] Design and implementation need not care much about the system compatibility and can build adaptive systems to ease the field system’s building and expanding,” Jianfeng said.
“Protocol means CPU overhead, and thus the data could be lost during the transmission. A Reflective Memory network transmits raw data with extreme latency, which means higher determinacy of data transmission and lower CPU overhead.” Reflective Memory monitors and duplicates data transparently and thus shares data without the software overhead, which is more cost effective as it eliminates the cost of additional development effort, testing, maintenance, documentation, and CPU requirements for traditional communication.
Could this be applied to all industrial environments? The answer is no. Reflective Memory, a unique GE real-time network technology, can be integrated with other GE embedded systems to build real-time systems to achieve remote data transmission. It can be used in all environments where computers or programmable logic controllers are connected using Ethernet, optical fiber channels, or other serial networks, such as flight simulators, telecommunication, high-speed progress control (rolling mills and aluminum factories), and high-speed testing and measurement systems. However, it is not applicable to all environments.
“Reflective Memory is more suitable to systems where real-time communication is top priority. Although the price of Reflective Memory is higher than that of hardware whose performance is low, it is rewarding considering its high functionality and usability,” Jianfeng said. Of course, he further pointed out that for those industrial environments where the requirement of real time is not demanding, it is fine to use a traditional network, considering cost. That is to say, “Reflective Memory is especially suitable to the environment where high-speed data transmission, real time, and determinacy are top priority. Up to 256 nodes can be connected to the ring, and that is enough for industrial environment and simulation applications.”
Real-time, accurately controlled production
Thanks to low latency and high determinacy, Reflective Memory is best applied to applications in metallurgy, steel, and communications environments, where real-time requirements are strict. For example, it can be used to improve the PLC performance to control the aluminum or steel rolling process. For an aluminum mill with 3500 ft/min speed, 2-3 feet of aluminum can pass through within the response time of the executor when using the usual PLC controller. The executer can apply or release pressure to roll out aluminum with various thickness. With Reflective Memory, data related to the mill can be input into the PLC, which writes the data to Reflective Memory. Thus data is sent to independent VME computers to transmit the complicated control logic algorithm. The system uses the simple Reflective Memory commands to transmit the output control data back to the PLC. The speed of data transmission and computer are so high that there is no delay in the PLC operation control loop.
“If there is any delay,” Jianfeng said, “the thickness of steel cannot follow the specifications, and some of the material will be wasted. Reflective Memory based on a VME advanced control system ensures real-time and accurate control to reduce the response time to as short as 4 in. [in this application] and thus improve product quality.
“Compared with PLC, Reflective Memory has better real-time performance, achieving high-speed system response with higher cost. Reflective memory is a good supplement to the traditional PLC control, and you can decide which solution to take. If there is some system redundancy, you can use Reflective Memory to back up the data in the time-out machine within a few microseconds, which is common in PLC control,” Jianfeng said.
- Sunny Jin (Jin Yan) is senior editor with Control Engineering China. This article appeared in an earlier edition of CEC and was edited by Mark T. Hoske, content manager, CFE Media, for use in Control Engineering, firstname.lastname@example.org.
Search Control Engineering China at www.controleng.com for other articles.
Real-time backup and fail-over of control for manufacturing information can augment reliability.
Applications include flight simulators, telecommunication, rolling mills, aluminum factories, and high-speed test and measurement systems.
High-speed data transmission, real time, and determinism are key attributes, for up to 256 nodes in a ring topology.