HP Fibre Channel Fabric Migration Guide

7
Fabric Migration Guide
Migrating to Fabric
the I/O is timed out and then retried for direct access devices. It
can also cause a loss of position for sequential access devices, and
the entire operation may need to be retried.
• Simpler link level protocol results in fewer chances of encountering
hardware, firmware, or software defects in the HBAs.
• A fabric environment can provide multiple paths of communication
between initiator and the target on the loop. If one path goes down
because of loop port/link problems, the switch can determine an
alternate path.
Manageability and Fault Isolation
• Switches contain management utilities which collect and present a
broad range of statistical and error information. This information
greatly assists in identifying specific components that are
experiencing errors.
• Link level faults are isolated to a single link segment on a fabric.
• Target devices that are directly connected to the switch no longer
need to have loop identifiers (Hard Physical Addresses) configured.
Performance
• Removal of FC-AL protocol for N_Port (DFA) devices.
FC-AL protocol overhead is related to frame size and the number of
members on the loop. FC-AL protocol can be quite significant for
larger loop configurations (loops with more than 30 active nodes).
When you move to a Direct Fabric Attach topology, the AL level of
protocol is no longer needed or present.
• Addition of fabric routing time.
Fabric routing times are generally less than 1 µsec (one microsecond).
For single switch fabrics or properly configured cascaded switches,
the fabric will have no noticeable effect on latency.
• Parallel transfer of data.
Fabrics allow parallel transfers of data to and from multiple initiators
on the fabric. They also allow parallel transfer of data to and from
separate targets at the same time with a single initiator.