Storage Basics: Fibre Channel, Part Two
In the first part of our two-part look at Fibre Channel, we covered the reasons that Fibre Channel has become popular as well as the standards and market challenges facing Fibre Channel. In this second part we look at implementation considerations such as topologies, switches, and port types.
Fibre Channel Topologies
Just like a network topology, there are accepted layouts that are used to define how Fibre Channel is implemented. The following sections describe these three accepted layouts.
In this, the simplest of all the Fibre Channel topologies, devices are connected directly to each other. An example of such a configuration would be a server connected directly to a storage array. In a point-to-point configuration, devices are connected directly to each other, allowing the communication between devices to use the entire bandwidth of the link. The simplicity of the point-to-point configuration makes it the cheapest way to implement Fibre Channel.
In a way, the Switched Fabric topology is similar to the star network layout commonly implemented on networks that use twisted pair cabling. A central device, in this case a fibre channel switch, provides a central connectivity point for all of the devices in the fabric. This allows each node to access all other nodes that are connected to the fabric.
The configuration of a fabric topology can be either cascaded or non-cascaded. In a cascaded configuration, all switches are interconnected to form a larger mesh. This allows a device connected to one switch to access any other node connected to a switch within the fabric. In a non-cascaded configuration, the switches are not interconnected, making this design easier to implement than the cascaded; however, the performance of the fabric will significantly degrade should a switch fail. Also, the mesh design of the cascaded fabric allows for greater fault tolerance.
There are several benefits of this fabric topology, including support for a large number of nodes (over 16 million in fact), increased fault tolerance when implementing a cascaded configuration, and improved scalability.
The Fibre Channel Arbitrated Loop, or FC-AL as it is known, is a Fibre Channel topology that combines the advantages of the fabric topology (support for multiple devices) with the cost savings of the point-to-point topology (where there's no need for a central switch). In a FC-AL implementation, devices are connected to a central hub, which is typically cheaper than a switch. As with Ethernet networks, the hub offers no additional functionality to the connected devices beyond serving as a centralized connection point.
To establish communication between nodes in an FC-AL topology, devices must take control of the loop and then establish a point-to-point connection with the receiving device. Once this point-to-point connection has been made, data can be transferred. Once the transmissions have ended, devices on the FC-AL can once again "arbitrate" to gain control of the loop for establishing subsequent point-to-point connections.
In the FC-AL topology, there can be up to 126 nodes connected to a single link, but as you might expect, the more devices that share the loop, the lower the performance. Because of this, it is highly uncommon to attach the theoretical 126-node maximum to a single link.