Impact of SAN multi-protocol routers (1)

In 1988, Indian Canadian Kumar Malavalli began his six-year-long creation of the Fiber Channel standard. With the unremitting efforts of him and other Engineers interested in this process, fiber channel was finally approved by the US National standardization administration as a national standard in 1994.

This great career of Kumar Malavalli originated from his comparison and Investigation of the world's popular network technologies represented by Ethernet and the channel technologies represented by SCSI. The basic starting point for creating the Optical Fiber Channel Protocol is to create an advanced network architecture that combines the advantages of the network technology and the advantages of the channel technology. Because Fiber Channel learned the network technology at that time, the advantages of the channel technology were eliminated, and engineers with various network technology backgrounds were competing to take the network they knew, the channel protocol was rewritten to the fourth layer (FC-4) of the fiber channel. At that time, the protocols that were standardized to FC-4 were SCSI, IP, ATM, FICON (ESCON's fiber channel version) and so on. Over the past 10 years, fiber channel SCSI has evolved into a mainstream protocol for storing Local Area Network SAN. FICON has also become the mainstream of the Mainframe (Mainframe) storage protocol. The IP Technology Based on Fiber Channel is applied to the management of fiber channel switches.

A fiber-channel-based SAN can transmit data at a high speed of 200 MB/sec. An important difference between an optical fiber channel and other network protocols is its data transfer bandwidth utilization. In an Optical Fiber Channel architecture, the bandwidth utilization can easily reach more than 99%. This is incomparable to other existing network protocols. Fiber Channel can extend the SAN connection distance to over 100 kilometers. If the protocol conversion technology (for example, the conversion from the fiber channel to the SONET and the conversion from the fiber channel to the IP address) is supplemented, the SAN connection distance can reach the global range. A Fabric consisting of fiber channel switches connected to each other can connect 239 fiber channel switches with up to 24-bit device address space. Compared with Ethernet, fiber channel data transmission in the same Fabric is device-to-device, while data transmission in the same Ethernet Subnet is broadcast. This is the main reason why fiber channel bandwidth usage is several times higher than that of Ethernet. It is precisely because the data transmission device of the Optical Fiber Channel in the same Fabric is opposite to the device, and the optical fiber channel stipulates a strict system to constitute a management system. In the management system of this system, the access and removal of devices, including fiber channel switches, are broadcast to devices that communicate with the access and removal devices in the form of broadcast.

SAN has brought many benefits to mankind in its historical development process, mainly in the following aspects:

SAN integrates server memory. This greatly improves memory usage efficiency and reduces system costs.

SAN enables centralized management of storage networks. Reduces storage network management costs. Improves management efficiency.

SAN frees the LAN network from the pressure of backup and access. Enable LAN to focus on business applications. This greatly shortens the backup access time.

The high redundancy of SAN has greatly improved the availability of the storage system.

SAN provides an ideal infrastructure with high redundancy and long transmission distance for Disaster Tolerance. SAN is one of the key elements for achieving high-speed recovery of disaster recovery systems.

Although Fiber Channel has completely changed the storage world, the rapid development of network storage and disaster tolerance has also gradually made people feel the limitations of traditional fiber channel. These limitations are mainly manifested in the following aspects:

A Fabric is formed by a vswitch that is connected to each other when the SAN network is formed by cascade of fiber channel switches. Data Interconnection and resource sharing cannot be achieved between different Fabric.

The more fiber channel switches in a Fabric, the larger the SAN network. There are more opportunities for Device Access and migration. As a result, more information is sent to the outbound broadcast. Although such broadcast information is insignificant compared with the data broadcast storm in Ethernet, it cannot be ignored for the high availability level and network management required by the storage network.

In various disaster recovery systems, once the local and remote SAN are connected, a large Fabric is formed. The long-distance bare fiber or IP connection between local and remote SAN is often the weakest link in this Fabric. On the premise that the local and remote SAN belong to the same Fabric, if the connection between them is unstable, Fabric reconfiguration that affects the whole SAN will occur ). This is an important cause of unstable disaster recovery systems.

The Disaster Tolerance System is at a historical development stage from traditional disaster tolerance between two points to mutual disaster tolerance between multiple data centers and the provision of disaster tolerance as a service to multiple customers. The traditional isolated Fabric Structure of fiber channel can no longer meet the requirements of Multi-Point Disaster Tolerance and make disaster tolerance a service.

Today, there are already hundreds of thousands of SAN islands in the world. Users often need to integrate these SAN islands. If this integration is integrated into a Fabric, the customer will face complicated operations such as adjusting the Fiber Channel switch parameters and rewriting the system composition files on some servers. In many cases, the customer cannot even arrange enough scheduled downtime to complete such system integration.

Different functional departments in the company objectively need to have the storage network development space and freedom of their own departments. The storage network of other departments should not be affected by the expansion of a department's SAN. However, it is difficult to achieve this when all the company's server memory is in the same Fabric.