Fibre Channel (Fibre Channel) protocol

First, the basic architecture of Fibre Channel

FC-4 Upper Layer Protocol:scsi,hippi,sbccs,802.2,atm,vi,ip

FC-3 Common Service

FC-2 Framing Protocol/flow Control

FC-1 Encode/decode

FC-0 media:optical or Copper,100mb/sec to 1.062gb/sec

Describe:

FC-0: Physical layer, customized to different media, transmission distance, signal mechanism standards, also defined fiber and copper interface and electrical

Cable Index

FC-1: Defining standards for encoding and decoding

FC-2: Defines frame, flow control, quality of service, etc.

FC-3: Defines common services, such as data encryption and compression

FC-4: Protocol mapping layer that defines the interface between Fibre Channel and upper layer applications, such as: Serial SCSI Association

, the HBA driver provides FC-4 interface functions, FC-4 supports multi-protocol, such as: Fcp-scsi,fc-ip,fc-vi.

Second, FCP-SCSI

FCP-SCSI: A serial protocol that maps a fibre channel device to a logical drive accessible by the operating system, which allows fibre Channel to be used without any modification to a previously SCSI application. FC-SCSI is the most important means of communication between storage systems and servers. The SCSI extension has the copy command, a new ANSI T10 standard, and also supports the direct movement of data between SAN storage systems through data migration applications.

The advantages of FCP-SCSI and bus connections have been proven on storage LANs, FCP-SCSI provides higher performance (100M/SEC), farther connection distances (up to 10 km per connection), and larger addressing space (maximum 16 million nodes). FCP-SCSI uses frame transfer instead of block transfer. Frame transmission transmits short, small transaction data in large data stream transmission mode, which improves service quality. FCP-SCSI supports the network configuration of the storage "pool" technology to simplify management and resource storage. FCP-SCSI supports coding techniques for improving reliability and availability.

Third, Fc-ip

FC-IP maps a Fibre Channel address to an IP address, FC-IP addressing by broadcasting an IP address and then returning a MAC address from the storage node. If SCSI devices cannot differentiate between FCP-SCSI frames and FC-IP frames, IP broadcasts can cause errors. The HDS system can distinguish between FCP-SCSI frames and fc-ip frames by detecting a frame header, and a storage system without this capability must prevent FC-IP frames from being broadcast to the fibre port by other means, such as switch zoning.

Fc-ip and Ethernet ratios have several advantages: they can be integrated with an internal connectivity architecture similar to FCP-SCSI storage to save on usage costs, faster and more efficient.

The Ethernet transport packet is up to 1500 bytes high. A packet is a basic calibration unit in an Ethernet that causes an interruption in the CPU cycle after each frame. Load is often a limiting factor in GB Ethernet, avoiding all bandwidth consumption. While the FC-IP data frame reaches 2000 bytes, the Fc-ip correction base unit is a multiple frame queue. The MTU can reach 64 frames, comparing Ethernet allows Fibre Channel to transmit more data between host interrupts. This MTU reduces the CPU cycles needed and improves transmission efficiency.

FC-IP also has the advantage of using Fibre Channel networks, which are closed networks based on flow control. The Ethernet was set up to take into account the public network without flow control, which, when blocking occurs, returns and contracts after a consistent period of time, consuming additional CPU cycles. IP applications run on FC-IP without modification, enjoying the high speed of fibre channel and greatly reducing processing disruption.

Emulex and JNI are the Fibre Channel HBA vendors that provide the FC-IP drive. They plan to pass a "Combo" to support FCP-SCSI and FC-IP. Troika provides support for FCP-SCSI, Fc-ip, Fc-ip (QoS) controllers, and QoS allows network administrators to assign protocol priority.

Four, Fc-vi

Fc-vi is a VI architecture on Fibre Channel that allows data to migrate quickly between the memory addresses of Fibre Channel contacts. Fc-vi is the VI architecture of the Fibre Channel application, a intel,compaq,100 many vendors and organizations in order to reduce server communication waiting for protocol standards. VI designed to achieve the cluster computer communication between the waiting for the reduction and high bandwidth effect. In a Fibre Channel network, through a method of direct access (DMA) between the buffer and the application memory of the HBA on the other node interface, this goal completes the VI schema to establish the memory registration mechanism, essentially restricting the memory address of the user's memory and supporting the data from the user's memory directly to the HBA's cache, This data can then be transferred via external media to the specified location (registration) of the other server's application memory. If VI is to be used, the application, database, or operating system must obtain the appropriate APIs from www.viarch.org. DB2 6.1 and Oracle8.1 both use the VI schema in their database cluster applications.

The latency of IP over Ethernet includes TCP stack (CPU load) and Ethernet transfer latency. The maximum transmission rate of 100BaseT 100MBIT/SEC,FC-IP reduces the Ethernet-related latency and transmits at the speed of Fibre Channel, providing better throughput than IP over Ethernet, but still avoids the software latency of TCP/IP. The FC-VI removes the TCP stack and provides DMA between the application memory and the HBA. Fc-vi the system kernel, avoids the operating system context conversion and buffering changes, and realizes a higher transmission rate.

FC-VI requires a Fibre Channel HBA,FC-VI HBA that supports the VI architecture and a Fibre Channel HBA that supports SCSI I/O is essentially different. Both Troika and Finisar provide Fibre Channel HBAs that support the VI architecture. Finisar sells a PCI based Fibre Channel HBA that supports the VI architecture, which supports point-to-point connections or switched forms. Troika to sell a PCI based intelligent controller? D? Dsan 2000 Series Controller, this controller supports FC-SCSI,FC-IP, point-to-point fc-vi,fc-al and switched topologies. The Troika controller provides a variety of management options and features, such as protocol priority configuration and the path transformation in load balancing.

V. Multi-Protocol architecture

All of these three protocols (FC-SCSI,FC-IP,FC-VI) can be assembled into a fibre channel structure. Although these protocols can also work in Fc-al, the corresponding bandwidth sharing and quorum consumption ignores the performance benefits of Fc-ip and FC-IP. Fabric swapping is recommended because the fabric Exchange provides access to multiple non-blocking 100m/sec between servers and between servers and storage devices.

HDS has developed a multi-protocol business intelligence solution that simulates the use of FC-SCSI,FC-IP,FC-VI switching architectures. Using Ancor, Brocade, or both of these fabric switches, the program was validated at the computer distributor exhibition in November 1999 and the Gartner Group Storage 2000 conference in June 2000. The fabric exchange in this scenario integrates centralized management of all data networks and provides unlimited scalability (with increased server and storage system nodes). The use of fabric swapping sacrifices some completely negligible performance losses, and in most cases FCP-SCSI consumes no more than 2% through the swap.

As more and more applications are modified or developed for the VI schema, we will see more and more data networks integrated into Fibre Channel. The emergence of market opportunities such as EAI (Enterprise application Exchange) needs to provide real-time data transmission and exchange between heterogeneous platforms and heterogeneous databases, Fc-vi makes it possible to exchange high-speed data between servers. Existing applications such as NAS can run their network file systems at the speed of Fibre Channel using VI sockets.

People are trying to access the storage of IP standards, Cisco for SCSI over IP to the IETF submitted a specification, the current specification is still in development, it needs to control and command signal and data signal transmission cable separate, mainly consider the flow control and transport control overhead.