The difference between PCI-5565 and PMC5565 reflective memory cards

Optical fiber reflective memory network key technologies:

1) Design and Implementation of High-Real-Time Low-Overhead Network Communication Protocol Currently, the mainstream open network protocols, such as TCP/IP protocol used in Ethernet, have complete functions and stable operation, but their transmission mechanisms and service methods are relatively complex and redundant. With poor real-time performance, it is not suitable for areas with high real-time requirements.

The optical fiber reflective memory network communication protocol should reduce the complexity of the protocol under the premise of ensuring high real-time performance and stability. Under the premise of ensuring the basic network service and transmission stability, try to improve the system's transmission performance, real-time performance and corresponding speed. At the same time, there must be a complete error handling mechanism to ensure that the error does not spread in the event of an error and that there is a good self-healing ability.

2) Data forwarding mode for edge-to-edge forwarding to achieve low-delay data forwarding Design of the protocol controller uses RFMMA (ReflectiveMemory Multiple Access) based on reflective memory multi-mode access technology to support data, IO, commands, interrupts and other Data transmission mode.

In the implementation of the protocol, that is, the design of the protocol controller, a low-delay data forwarding strategy is adopted, and the forwarding delay is controlled within 1us, which poses a high challenge to the implementation of the protocol controller. The following methods are mainly used: (1) Adopting the data transmission mode of edge-edge forwarding, not adopting the store-and-forward mode, reducing the forwarding delay of data on a single node, and improving the real-time performance of the system. (2) In the design of the protocol, the design is optimized, the length of the information header is compressed, and important information is placed in front of the information header to facilitate rapid judgment when forwarding. (3) A complete data frame recovery mechanism has been designed to ensure the reliable recovery of waste data frames by means of node IDs and transmission counters. At the same time, the CRC checksum is used to verify the correctness of the data.

Generally, network interface devices use the storage and forwarding mode for forwarding data, and the forwarding delay is too large, which leads to an increase in the delay of the entire system and reduces the real-time performance. The use of instant receive / transfer (edge-to-edge forwarding) mode, to complete the processing and forwarding at the same time when receiving data, instead of the common store-and-forward mode, only 4 clock cycles to complete the judgment & forwarding, the typical data frame is (length 256Byte ) Forwarding delay is more than 30 times shorter than store forwarding. The measured delay is less than 0.6us.

3) Network Hub Dynamic Reconfiguration and Data Monitoring Technology of Fiber HUB Utilize high-speed switch arrays and high-speed transceivers integrated with FPGAs to build a switch array. Through the switching of switch arrays, data transmission paths for both inner and outer rings are formed. A circular transmission method is formed logically, and data is collected in real time by an FPGA transceiver to implement network status monitoring and self-diagnosis of fault nodes.

4) High-speed circuit design and simulation verification technology The baud rate of optical fiber reflection memory network is 2.5Gbps. This data transmission rate is based on this data communication rate, which is a high-speed circuit design. In order to ensure the reliable transmission of data and guarantee the system's electromagnetic compatibility and signal integrity, higher requirements are placed on the PCB design of the circuit board. In order to ensure the design quality, Cadence's SigXplorer software was used to simulate the circuit and multiple iterations of the simulation results were used to improve the design.

Reflective memory strong real-time star network description:

Reflective memory real-time networks are the only high-performance star networks that are specifically developed for applications requiring real-time, stable multi-processor data transmission. The reflective memory star network structure has congenital advantages over the ring network in terms of stability. The failure of one node only affects itself and the other nodes work normally. Reflective memory star network provides enhanced network reliability, high bandwidth, and low latency, and is the best solution for emergency mission applications. High priority can be assigned to important nodes to ensure the lowest latency. Provides full-featured, easy-to-use software that supports multiple operating systems (Windows, Linux, Vxworks, IRIX). Software includes APIs to maximize the portability of convenient code and reduce application development costs and time.

Reflective memory strong real-time star network works:

Reflective Memory A real-time network provides a memory area that is shared by all nodes at the same time, and each node has a physical reflection of that memory area. Data written to a node's reflective memory space will be sent to all other nodes at the same time. The reflective memory real-time network merges data from different nodes into unified data and updates all nodes. The reflective memory real-time network ensures that data arrives at all nodes at the same time and maintains a consistent byte order. Just by accessing the node's own reflective memory, all nodes can use interrupts, send data, and messages transparently and steadily.

Reflective memory real-time network advantages:

At the same time accepting data from each node without the memory-mapped pointer of the "C" language for the ring-delay network problem, which facilitates high throughput, low latency, and high transfer rate (1.0625 Gbits/s), ensuring the fastest speed update. All nodes update all node data regardless of the number of nodes, and there are essential differences between the ring network and ease of use, maintenance, development and easy to increase nodes to reduce cable complexity comprehensive software support, provide advanced diagnostic software network support node interrupt detection and notification capabilities. The failure of one node only affects itself, and other nodes work normally and quickly isolate the failed node and cable reflection memory card (PVIC) signature data simultaneously to all nodes. 1.0625 Gbits/s transmission rate Updates all nodes in consistent byte order Low access latency When (20 nsec) flexible memory address space maps to allow direct access to remote memory for each node and the cable works alone, if one node is damaged, the other node continues to work supporting up to 500 meters of fiber-optic cable 66 MHz/64-bit PCI/CompactPCI /PMC comprehensive error monitoring and error recovery capabilities using memory-mapped pointers, in conjunction with standard APIs, to eliminate programming complexity. Each card supports 2 programmable interrupts, 7 semaphores and 2 PCI cards can be directly connected to 192K words on the board Section 32-bit dual-port memory supports sharing without the need to access remote memory drivers supporting multiple operating systems (Windows, Linux, Vxworks, IRIX) to provide exclusive development of Dos drivers.