SAN Introduction

SAN Architecture

Interface

Fibre Channel Concepts

Topologies

Hardware Components

Environmental Concerns

Software Components

Fibre Channel: Topologies

The Fibre Channel standard supports three principal topologies: point-to-point, arbitrated loop, and switched fabric. These topologies are analogous to the bus, ring, and star topologies encountered in traditional data networks. Each topology has unique features, advantages, and disadvantages that will be discussed.

Overview

Let's take a quick look at each of the topologies. You can click on the small graphic next to each topology type and scroll down for additional information.

Point-to-Point: this is the most basic of the three topologies. It's used to connect two devices together-- a server and a storage device, for instance.

Arbitrated Loop: Capable of connecting up to 126 devices, the Arbitrated Loop topology is more robust and provides better scalability than the Point-to-Point topology.

The loop's total bandwidth is shared among all of the devices included in the fabric.

Switched Fabric: Through cascading, the Switched Fabric Loop presents even greater scalability than the Arbitrated Loop topology.

Using switches, this topology routes device traffic automatically to efficiently distribute bandwidth.

Arbitrated Loop
The Fibre Channel Arbitrated Loop provides scalability, allowing up to 127 nodes (126 loop ports plus one fabric port). The total bandwidth of the FC network is shared among all of the devices on the loop; each device must request control of the loop before transmitting data. To request control of the loop, a node sends a special signal called a primitive. If the signal is returned to the initiating node with that node's address, control of the loop has been granted. This arbitration method is similar to the token-passing scheme used in Token Ring data networks. However, unlike with a Token Ring, once the device has successfully arbitrated, it has control of the loop for as long as it needs.


Architecture
The Arbitrated Loop topology is self-discovering. The NL_Port on each node allows the node to discover its environment and begin working with other nodes on the loop without manual intervention. If the loop is not attached to an FC switch, it is called a private loop. Attaching the loop to an FL_Port on a switch creates a public loop. Only one FL_Port can be attached to a loop, and a dedicated address (independent of the 126 NL_Port addresses) is reserved for this connection.


Addressing
In addressing, the process of deciding which port takes precedence in communicating is called contention. To resolve contention, every port requires a unique physical address. This procedure is called Arbitrated Loop Physical Addresses, or ALPA. Under ALPA, the lower addresses generally get to "go first".


Arbitrated Loops in Dell Products
Fibre Channel Disk Processor Enclosures (DPEs) and Disk Array Enclosures (DAEs), such as the Dell PowerVault 650F/651F and PowerVault 630F/631F, incorporate arbitrated loops into the disk backplane.

A Port Bypass Circuit (PBC) is used in the backplane to detect the presence of an active disk node on each available port. The PBC provides hot-swap functionality by opening the loop when a new disk is inserted, or closing the loop if a disk fails or is removed. Fibre Channel hubs also incorporate PBCs to allow star wiring and online scalability of an FC-AL network.

Switched Fabric

A fabric incorporates Fibre Channel switches to provide a network with non-shared bandwidth (each device within the fabric has its own dedicated connection). The switches provide for the greatest level of scalability of the three topologies. Switches can be cascaded via an inter-switch link (ISL) to increase the available number of nodes.

Theoretically, a fabric can support 16 million nodes.

Configuration

The self-discovery and self-configuration features of Fibre Channel greatly ease the management of a switched fabric FC network.

Universal ports on FC switches, such as those on the Dell PowerVault 51F and PowerVault 56F, can automatically configure themselves as F_Ports, FL_Ports, or E_Ports.

The auto-configuration works like this:

If a universal port on a switch is connected to an N_Port on a node, the universal port works as an F_Port.
If the switch port is connected to an FC-AL loop, it acts as an FL_Port.
If the switch port is connected to another switch, it acts as an E_Port and forms the inter-switch link between the two switches.

When a node is connected to a switch, it will automatically log in and determine how to interoperate in the fabric. Each switch port is assigned a 24-bit address. When a node's N_Port is connected, it inherits the address assigned to the corresponding F_Port on the switch. When a loop is connected to an FL_Port on the switch, it creates a public loop by inheriting the first 16 bits of that port's address, and using the remaining 8 bits to provide addresses for the members of the public loop. Of the 128 possible addresses, one is reserved for communicating with the switch's FL_Port, and one is reserved for the FC-AL loop itself. As a result, 126 nodes may be attached to the public loop.

Advantages

Limitations

With fabric, devices theoretically have a dedicated connection.
Greater scalability than arbitrated loop.

More expensive than the arbitrated loop topology.
More complicated than arbitrated loop in planning and implementation.

Fibre Channel: Topologies

Overview

Point to Point

Arbitrated Loop

Architecture

Addressing

Arbitrated Loops in Dell Products

Switched Fabric

Configuration