Chapter 3: Subsystem Connection
Topology and
Configuration Considerations
3-5
3.2.2 Fibre Channel Topologies
The Fibre Channel standard supports separate topologies. They are Fibre
Channel Arbitrated Loop (FC-AL) and fabric switch topologies.
•
FC-AL:
This is the most common topology currently in use. Fibre
Channel devices are all connected in a loop. Each device is
assigned an arbitrated loop physical address (AL_PA). The FC-AL
supports 124 devices in a single loop.
•
Fabric:
The fabric topology supports up to 2
24
Fibre Channel
devices. This topology allows many devices to communicate at the
same time with dedicated bandwidth. A Fibre switch is required to
implement this topology.
3.2.3 Host-side Topologies
The primary concern for configuring host-side topologies is to eliminate
points of failure. It is therefore recommended that the host interfaces be
connected to dual-redundant cabling devices, e.g., HBAs. It is also preferred
to connect the subsystems to host computer(s) through intelligent Fibre
Channel switches that feature zoning and fault recovery algorithms.
NOTE:
To create fault-tolerant, dual-redundant data paths on the host side, it is
necessary for a multipathing software to be installed on the host
computers to manage the I/O paths.
3.2.4 Drive-side Connection
The drive-side connection is made through redundant data paths on the
back-end PCB. I/Os are dynamically distributed through these separate data
paths; and, in the event of single RAID controller failure, the surviving
controller can access disk drives through alternate data paths.
3.2.5 Internal Connections
The two (2) RAID controllers are connected to each other internally through
a common backplane. With dedicated communication channels strung
between the RAID controllers, their caches can be synchronized and
workload can be dynamically distributed according to firmware load-
balancing algorithms.
