What is the architecture of a SAN network?
1. The protocol used in Gig-E is TCP/IP over Ethernet. The protocol in a SAN is SCSI over Fibre Channel. Fibre Channel uses frames with a payload of 2112 bytes that provide better throughput than Ethernet, and the stack used in Fibre Channel is more efficient than the normal TCP/IP stack as there are fewer layers to go through. Fibre Channel uses a mechanism called an Exchange as the method for two end nodes in a SAN to coordinate how they will communicate. Once communication is established, Fibre Channel can group related frames into a sequence to transmit large blocks of data. SCSI is a block-based protocol that is used for communication between applications and disk drives. Fibre Channel provides the underlying physical and data link layer to make this happen. You can run IP over Fibre Channel if you want. You can also run things like VI for high-speed inter-server communication.
2. SANs are built into "fabrics" to enable servers to communicate with storage devices. Ethernet is used for communications between servers. Ethernet uses a MAC address to locate end nodes through the address resolution protocol (ARP) Fibre Channel uses a WWN for addressing. You can have just over 16 million addresses in a Fibre Channel fabric.
3. Where Ethernet usually runs over twisted pair copper cables, today's Fibre Channel standards make use of 50u optical cables to transmit serialized SCSI at the speed of light. Although Fibre Channel can run over copper, that method is not used much anymore.
4. Fibre Channel switches use a different name server method and a different though similar method of routing data through the network.
5. SANs usually are built as two separate "fabric" networks for redundancy and each host uses two adapters to connect to the SAN.
Editor's note: Do you agree with this expert's response? If you have more to share, post it in one of our .bphAaR2qhqA^0@/searchstorage>discussion forums.
This was first published in September 2002