Optical fiber
The full name of the optical fiber is called optical fiber. The English name is OPTICFIBER, or OPTICALFIBER. It is made of pure quartz and drawn into filaments by a special process. The diameter of the optical fiber is larger than that of a human hair. The silk is even thinner. The characteristics of optical fiber are: fast transmission speed, long distance, large content, and no electromagnetic interference, not afraid of lightning, difficult to eavesdrop on the outside, non-conductivity, no trouble of grounding between equipment, etc.
Fibre Channel
In high-end server/workstation hard disks, Fibre Channel will also be used as the SCSI hard disk interface. Fibre Channel is a high-performance connection standard for two-way, serial data communication between servers, mass storage sub-networks, and peripherals through hubs, switches, and point-to-point connections. For large amounts of data that need to be efficiently transferred between servers and storage media, Fibre Channel provides remote connections and high-speed bandwidth. It is an ideal technology suitable for storage local area networks, cluster computers and other data-intensive computing facilities. Its interface transmission speed is divided into 1GB and 2GB and so on.
1. The origin of Fibre Channel technology The explosive growth of data in the information age provides a good opportunity for the development of storage technology. Now information executives are more concerned about how to Safe storage, management and use of data. Therefore, people not only have higher and higher requirements for storage device capacity and performance, but also put forward technical requirements for storage systems with high performance, high reliability, and long-distance transmission. Fiber Channel (FiberChannel) technology was born under the drive of this demand. At present, in the design of storage systems, all business systems that involve operating on large relational databases and reading massive amounts of data generally tend to adopt Storage Area Networks (Storage Area Networks,) architecture. The storage area network (hereinafter referred to as "SAN") is a network system based on the networked I/O storage protocol that enables "anytoany" connection and communication between servers and storage devices. The development of SAN drove the development of Fibre Channel technology, and the development of Fibre Channel architecture paved the way for the technical concept of SAN. Fibre Channel technology is a protocol architecture based on Fibre Channel, which started in 1989 and formulated the corresponding ANSI standard in October 1994. In addition to optical cables, the transmission medium of fiber channel technology also has other transmission carriers such as copper cables, but it is usually called optical channels in the world. Fibre Channel technology can be rapidly developed and widely used (reflected in the large number of mainstream SAN systems using FC technology), not only because Fibre Channel has higher bandwidth, longer connection distance, better security and scalability, More importantly, Fibre Channel technology combines the advantages of channel technology and network technology. Using Fibre Channel networks can create a storage area network (SAN) that is different from the familiar local area network (LAN) or even the metropolitan area network (MAN). SAN is not a product, but a method of configuring networked storage. The main idea is to convert the data exchange on the traditional network to a SAN mainly composed of storage devices and database servers. With the help of Fibre Channel technology, SAN supports long-distance communication, and completely separates data storage and application services, so that storage devices can become shared resources that can be accessed by all servers connected to the SAN at high speed, security, and reliability. At the same time, the SAN also allows each Storage devices, such as disk arrays and tape libraries, can work together without going through a dedicated intermediate server. SAN solves the problem that once a large amount of data access occurs in the traditional LAN, the network performance will be greatly reduced, so that data access, backup and recovery will not affect the performance of the LAN, fundamentally guarantee the service quality of the application system, and can greatly To reduce administrative expenses. 2. Fibre Channel Protocol and Hierarchical Model Fibre Channel is a technical standard. It is a common name for a set of integrated standards jointly developed by several committees commissioned by the American National Standards Institute (ANSI). It is used to improve multi-disk storage systems. High-performance interface standards designed for speed and flexibility. It is independent of the medium and supports simultaneous transmission of multiple different protocols, such as IPI, IP, FICON, FCP (SCSI) and other protocols. It is suitable for servers, mass storage sub-networks, and peripherals for bidirectional, Serial data communication. Just as in Ethernet, protocols such as IP, NetBIOS, and SNA can be used on a single Ethernet adapter at the same time, because all these protocols are mapped in Ethernet, and various network layer communication protocols can also be mapped through protocols Realized on Fibre Channel. The advantages of Fibre Channel technology are mainly reflected in: (1) high bandwidth, currently 200MB/s data transfer rate has been achieved, 400MB/s has passed the test; (2) high-capacity addressing capability and capacity expansion capability, which can access 16 million nodes(3) High data concentration and global sharing of storage capacity; (4) Long connection distance between each pair of nodes, multi-mode optical cable up to 500 meters, single-mode optical cable up to 10 kilometers; (5) modular expansion and connection Method: (6) Use fiber optic switches and related software to establish a highly available or fault-tolerant service system; (7) can facilitate the establishment of load balancing and server cluster systems. Fibre Channel technology is a new technology developed by combining the advantages of "channel technology" and "network technology": channel technology is a hardware-intensive technology because it is designed to quickly transmit a large amount of data in the buffer zone. Connect devices directly without using too much logic; network technology is a software-intensive technology, because data packets need to be routed to a node in many devices on the network, and network technology has the ability to operate a large number of nodes. Fibre Channel technology combines the above-mentioned advantages of channel technology and network technology from the beginning of its design. Five independent levels are defined in the Fibre Channel protocol, from the physical medium to the high-level protocol transmitted in the Fibre Channel, which contains the full picture of the Fibre Channel technology. The following are the functional modules of these five layers: ①FC-0, the physical layer, defines the physical port characteristics of the connection, including the physical characteristics, electrical characteristics, optical characteristics, and transmission rate of the medium and connectors (drivers, receivers, transmitters, etc.) And some other connection port characteristics. The physical media are optical fiber, twisted pair and coaxial cable. This layer defines how light is transmitted on optical fibers and how the transmitter and receiver work on various physical media.②FC-1, transmission protocol, FC-1 according to the ANSIX3T11 standard, stipulates the 8B/10B encoding method and transmission protocol, including serial encoding, decoding rules, special characters and error control. The transmission coding must be DC balanced to meet the electrical requirements of the receiving unit. Special characters ensure that what appears in the serial bit stream is a short character length and a certain transition signal for clock recovery. This layer is responsible for obtaining a series of signals and encoding them into usable character data.③FC-2, frame protocol, defines the transmission mechanism, including frame positioning, frame header content, usage rules, and flow control. The length of the fiber channel data frame is variable and the address can be extended. The fiber channel data frame length used to transmit data is up to 2K, so it is very suitable for the transmission of large-capacity data. The content of the frame header includes control information, source address, destination address, transmission sequence identification, and switching equipment. The 64-byte optional frame header is used for protocol mapping when other types of networks are transmitted over Fibre Channel. Fibre Channel relies on the content of the data frame header to trigger operations.④FC-3, public service, provides public services with advanced features, that is, the structure protocol and flow control between ports. It defines three services: Striping, HuntGroup, and Multicast. The purpose of striping is to use multiple ports to transmit in parallel on multiple connections, so that the I/O transmission bandwidth can be expanded to the corresponding multiple; the search group is used for multiple ports to respond to the same name address, it passes Reduce the probability of reaching a "busy" port to improve efficiency; multicast is used to deliver a message to multiple destination addresses. ⑤FC-4, the protocol mapping layer, defines the mapping relationship between the bottom layer of the fiber channel and the upper layer protocol (UpperLayerProtocol) and the application interface with the current standard. The current standard here includes all existing channel standards and network protocols, such as SCSI Interface and IP, ATM, HIPPI, etc.? It can be seen that the Fibre Channel protocol stack is the transmission carrier of a variety of high-level data protocols, especially for the transmission of SCSI and IP data. The process of transmitting the high-level data protocol as a carrier is actually a process of mapping the high-level data protocol to the physical layer transmission service of the protocol stack. Among them, the most commonly used fiber path protocol (FibreChannelProtocol) is the mapping of SCSI data, commands, and status information to FC physical layer transmission services. FCP has the independence of working on all optical fiber path topologies and all types of services.The following are the protocols mapped to the fiber channel: ①Small Computer System Interface (SCSI), that is, the mapping of the SCSI-3 protocol of the Fiber Path Protocol (FCP), is the main protocol mapped to the fiber path.②IP protocol.③Visual interface structure (VIA).④High-performance parallel interface (HIPPI).⑤IEEE802 logical link control layer.⑥Single-byte instruction code set (SBCCS), SBCCS is the realization of the instruction and control protocol in the ESCON storage I/O path used in the IBM large-scale system.⑦ Asynchronous transfer mode adaptation layer 5 (AAL5).⑧ Fiber Optic Connection (FICON), FICON is an upper layer protocol that maps the ESCON network communication protocol in the IBMS/390 host architecture to an optical fiber path network.1. The physical layer of the Fibre Channel network The physical layer of the Fibre Channel network is composed of the following three basic physical units: (1) Port: the interface used to connect the server system and the optical switch, or the interface used to connect the storage device and the optical switch. (2) Network equipment: optical fiber switch that uses optical fiber protocol for communication. (3) Cable: Used for the connection between the server interface and the optical switch interface, or the connection between the interface of the storage device and the optical switch interface.2. Network name and address elements The basic elements of network names and addresses in optical fiber networks are as follows: global name, port address, arbitrated loop physical address, and simple name server.(1)Global name The global name WorldWideName (WWN) refers to an 8-byte identifier assigned to each product, which can be used for a port in an optical fiber network. WWN is stored in non-volatile memory, and its format is defined by IEEE to provide a unique identification for each product in its installation network. When a node first logs on to a switch, it can exchange a full WWN of the N port with the switch. If there is no information about the N port on the switch, there will be a registration process, during which the N port sends Its own information is given to the switch, and the switch puts this information in its simple name server so that other processes and applications can access it.(2) Port address There are two types of port addresses in optical fiber networks: fixed addresses and dynamic addresses.①Fixed address: Each fiber channel identifiable device has a fixed fiber channel address, which is similar to the MAC address owned by each Ethernet card. This fixed address is unique in the world, and other devices can access it through this address.②Dynamic address: To support high-level addressing, Fibre Channel defines a 24-bit dynamic identification address in the Fabric domain. Each N_Port has a unique 24-bit N_Port identifier in the Fabric domain. N_Ports can either obtain its pre-set N_Port identifier through the protocol, or it can be dynamically allocated by the Fabric when the device logs in.(3) Arbitration ring physical address The arbitration ring physical address (ALPA) is a single byte, which uniquely identifies each port on the ring network. Each port in the ring network stores the addresses of all other ports in the ring, thereby providing a mechanism for communication in the ring. The port address can be used to determine whether a port on a ring is public or private.(4) Simple Name Server Simple Name Service provides a thin directory service. Nodes, switched fiber optic networks, and applications obtain port access information by using simple name services.3. Service level service level defines which mechanism is used in data transmission, and different service levels are used for different data. Service levels are divided into five categories: Level 1: Connection-oriented service with confirmation; Level 2: Connectionless service with confirmation; Level 3: Connectionless service without confirmation; Level 4: Connection-oriented part of the bandwidth Service; level F: communication format between switches.Flow control is a mechanism defined in the service level, which is divided into end-to-end flow control and buffer-to-buffer flow control. (1) The end-to-end flow control is that the receiving port transmits a return frame to the sender to confirm the receipt of the transmission frame; when the sender receives the feedback of the acknowledgement frame (ACK), it will set the credit value to 1, so that The next frame can be sent. (2) The flow control from the buffer area to the buffer area is a mechanism used between node ports of fabric ports or between two node ports to ensure that the device can receive the maximum number of frames. A R-RDY (receiver ready) primitive signal is sent out, indicating that the receiver can accept the frame; if the receiver sends a certain number of R-RDY signals, it means that it has enough buffer space to receive this number of frames . In addition to flow control, the service level also indicates whether the connection is dedicated. For a connection-type transmission process, a frame that is not delivered to a dedicated receiver address cannot be sent. In addition, it is not possible to send frames that are not of the same level in a certain level, so as to ensure that the connection can use the full bandwidth.4. Port type All components (ie, devices) in the Fibre Channel network use ports as network connections. The ports in the fiber channel network include the following basic types: N-port port, F-port port, L-port port, NL-port port, FL-port port, E-port port, G-port port. Among them, N, L and NL ports are used for terminal nodes in the Fibre Channel network, and F, FL, E and G ports are realized in Fibre Channel switches. ①N-port port and F-port port The initial Fibre Channel network includes two types of ports: one is the network port of the N-port port; the other is the exchange of the F-port port Fiber optic port. The N-port port is to access the storage devices on the Fibre Channel network and the port on the computer system. The task is to initialize and receive frames. If there is no N-port port, there will be no data communication on the network; the F-port port is optical fiber The port on the switch is used to provide management and connection services on behalf of the N-port port. These services are provided for the communication between each pair of N-port ports (host system and storage device). There is a one-to-one relationship between N-port and F-port. On the fiber switch in the fiber storage area network, there is only one N-port port connected to the F-port port. The communication between the other N-port ports and the N-port port in the fiber channel network is through their respective switches. The port initialization process on the computer and the communication of the N-port port are realized. No matter whether the N-port port is sending or receiving data, it always communicates with the F-port port. When there is no data transmission, the N-port port sends an IDLE frame to the corresponding F-port port on the switch to establish a "heartbeat" between the N-port port and the F-port port, so that the possibility can be detected quickly The connection problem occurred.②L-port port The L-port port exists in the Fibre Channel ring network. Unlike switched networks, nodes in a ring network share a structure of cable bandwidth. Similar to the N-port port in the switched network structure used for initialization to communicate with the F-port port, the L-port port is designed to initialize direct communication with other L-port ports in the ring. However, there is no port name corresponding to the F-port port in the optical fiber ring network. Because the optical fiber ring network is a logical ring, it is designed to work in an environment without a network hub. Therefore, if it is not required, the hub cannot provide the established port function for the ring network. The hub in the optical fiber ring network only plays the role of connection and prevention of failure. ③NL-port port and FL-port port When the Fibre Channel loop is added to the Fibre Channel network, communication between the N-port port node and the L-port port node must be allowed, which is defined Two new ports have been added: FL-port port and NL-port port.FL-port is a port on a fiber optic switch, allowing it to be added as a special node in a fiber channel network. The fiber channel ring network reserves only one address for the FL-port port, that is, it is impossible for two fiber switches to communicate at the same time. NL-port port is a port located in the ring network. It has the dual capabilities of N-port port and L-port port. It supports switched optical fiber network and optical fiber ring network at the same time, thereby enabling the communication between switched optical fiber network and optical fiber ring network. It became possible.④E-port port and G-port port In fiber optic switches, there are two common ports, they are E-port port and G-port port. G-port port is a "universal" port, which can be used for different ports in the switch such as F-port port and FL-port port. E-port port is a special port used for cascading fiber switches. The above are various ports that can be encountered in a Fibre Channel network. The fiber optic switch we use in the storage platform of the Ministry of Land and Resources is the Brocade fiber optic switch. The ports of this fiber optic switch support self-configuration. The self-configuring port can detect the port mode of the other end of all connections, and automatically configure it to support the mode of operation.5. Cables and media Many characteristics of SAN are determined by the physical layout of the network. The type of media selected in the SAN will affect the scalability and functionality of the SAN. There are two choices of media type: copper core wire and optical fiber.①Copper core wire The advantage of copper core wire is that it is the cheapest medium for connecting SAN components. The copper core wire is usually a 150 ohm copper core twisted pair. The transmission rate of copper core wire is 100MB/S gigabit transmission, and its effective transmission path is 0 to 25 meters without any attenuation. Both ends of the copper core wire usually use HSSDC connectors or DB-9 male connectors.②Multimode fiber The diameter of multimode fiber is usually 50 and 62.5 microns, and there is no difference in speed between them. The wavelength range of multimode fiber is 850 nanometers and 1300 nanometers. Light with a wavelength of 850 nanometers is visible and harmless to human eyes. The 1300 nm wavelength is invisible and harmful to the retina. There are many types of multi-mode fiber end connectors, including SC, LC and MT-RJ. Multimode fiber uses a concentrated LED rather than a real laser.③Single-mode fiber Single-mode fiber is suitable for long-distance signal transmission. Its wavelength is 1300 nanometers, which is invisible and harmful to human eyes. The diameter of a single-mode fiber is 9 microns. Because its diameter is so small, when using it for long-distance signal transmission, the light waves are not easily changed. So in the long-distance SAN, single-mode fiber is the best solution. Due to the small diameter of single-mode fiber, its potential launch speed is also the highest. The theoretical limit speed is 25Tb/s, while the theoretical limit speed of multimode fiber is 10Gb/s. Single-mode fiber itself is not much more expensive than multi-mode fiber or copper core wire. The increase in price is mainly due to its transceiver components, because it uses lasers instead of LEDs. Because the diameter of the single-mode fiber is very small, the accuracy of the fiber optic transceiver is very high.④Optical fiber connector There are many types of optical fiber connectors. In actual use, as long as the connection is clean, the use of the type of connector will not have any effect on the performance. The number of connections should be minimized when building a SAN, because light will be reflected back and forth between poor-quality connections in its path equipment. Therefore, the fewer the number of connections, the lower the probability of false signals in the SAN. The copper core connector used in many HBA (fiber interface cards, inserted in the PCI slots of the server system) card is HSSDC copper core connector.