There are three types of Fibre Channel connectivity; Point to Point, Fibre Channel Arbitrated Loop (FC-AL) and Switched Fibre Channel. Point to point fibre channel is simply used to connect two devices together, and so is not used to construct a SAN. The other options are;
FC-AL
FC-AL essentially daisy chains all the attached storage devices onto a fibre channel. If a server connection is lost, then the whole FC loop is lost, unless you use a hub. If fibre channel nodes are cabled through a hub, the hub heals the loop in the event of node failure, bypassing the non- operational node and passing signals directly between operating nodes. All the devices on the loop share the bandwidth, so the bandwidth on the fibre is divided by the number of devices attached to it. When a device wants to send IO, it first passes an arbitration package around the loop, to make sure the loop is available. If the package gets right round the loop without a collision, the sending device can then start passing data to the receiving device. The Fibre Channel loop is then reserved until the communication is complete, and there is no limit on how much data can be passed in one operation. Other senders must wait for the loop to become free.
An FC-AL loop can only support up to 128 devices. When a new device is added to the loop, all the device will probably change their physical addresses to cope, and this can cause problems for any attached server operating systems.
On the plus side, FC-AL with FC-hubs is cheap. It is often used to attach tape devices to several servers, and is typically used inside storage subsystems to manage the disk strings.
Switched Fabric
Unlike FC-AL, switches do not share bandwidth. They provide a virtual, dedicated connection between 2 points, so every connection gets the full bandwidth, typically 2 or 4 Gb/s. This is called a non-blocking architecture, and also means that several servers can communicate with one storage device at the same time. Switched fabric is reasonably priced, and easy to manage for medium sized networks. However, all your devices must support switched fibre channel. Switches are connected together using Inter Switch Links (ISLs) to create a fabric. This means that fewer ports are available for direct connection, and large configurations can be quite complex. The ISLs are not shown in the diagram above, to keep it simple.
A switched fabric is more scalable than FC-AL In theory it can support up to 16,000,000 connections, and can also support arbitrated loop devices.
Big switches are called directors, but the distinction between the two is becoming increasingly blurred Directors have more ports than switches, and have more built in redundancy, typically providing "5 9s" resilience. Large numbers of components can be connected with fewer ISLs, making the network less complex. As fewer ISLs are needed, a higher percentage of ports are available for device connection. The downside is that directors are much more expensive than switches
Core / Edge design fabrics are becoming popular, where the core is a high performance director, which is connected directly to high performance servers and storage, while appliances that need lesser performance are connected to the core by slower edge switches.
Federate Fabrics contain redundant switches, so that every server is connected to two switches, and has two independent paths through the SAN to the storage. The host servers must have multi-pathing software that can automatically failover if a path fails, and ideally load balance when two paths are available.
A multi-fabric SAN has two or more completely separate fabrics for redundancy, so an entire fabric can fail, and applications will still work through an alternate fabric. This is obviously expensive, both in terms of initial cost and in the management effort required to ensure that both fabrics are always identical.
iSCSI SANS
iSCSI SANs were proposed at the end of 2002 and matured into usable products in 2004. The biggest advantage of iSCSI is that it is much cheaper than fibre channel. It also makes clustering easier, supports multi-path IO and makes multi-site replication easier. ISCSI is still considered the optimum choice for moderate to low performance applications, though many people do use it for top class work. It is typically used for Windows, Netware or Linux, with some small Unix takeup. Application servers that are Ethernet enabled can be attached directly to an iSCSI SAN, they do not need expensive HBA cards.
An iSCSI solution provider can be found in iSCSI
SAN
iSCSI storage solutions developed by IP SAN Company SANRAD: Secure and manageable, SANRAD's iSCSI and IP SAN solutions allow you to manage your IP storage network easily and effectively.
This site includes some excellent white papers on iSCSI usage.