Specifications

DATA CENTER BEST PRACTICES
SAN Design and Best Practices 56 of 84
•Implement serial console server: Implement serial remote access so that switches can be managed even
when there are network issues or problems during switch boot or rmware upgrades.
•Use aliases: Use “aliases,” which give switch ports and devices meaningful names. Using aliases to give
devices meaningful names can lead to faster troubleshooting.
•Congure supportftp: Congure supportftp for automatic le transfers. The parameters set by this command
are used by supportSave and traceDump.
•Congure ntp server: To keep a consistent and accurate date and time on all the switches, congure switches
to use an external time server.
Firmware Upgrade Considerations
Both xed-port and modular switches support hot code load for rmware upgrades.
•Disruptive versus non-disruptive upgrades:
•Directors versus switches
•Simultaneous upgrades on neighboring switches
•Standard FC ports versus application and special-feature ports
•Review the Brocade Fabric OS Release Notes for the following:
•Upgrade path
•Changes to feature support
•Changes to backward compatibility
•Known issues and defects
•Consider a separate AG rmware upgrade strategy. Brocade Access Gateways have no fundamental
requirement to be at the same rmware release level as Brocade FOS. Upgrading only directors and switches
minimizes the infrastructure changes required during an upgrade cycle.
NPIV AND THE BROCADE ACCESS GATEWAY
One of the main limits to Fibre Channel scalability is the maximum number of domains (individual physical
or virtual switches) in a fabric. Keeping the number of domains low reduces much of the overhead typically
attributed to SAN fabrics. Small-domain-count fabrics are more reliable, perform better, and are easier to
manage. You can reduce overhead by doing the following:
•Reducing inter-switch zone transfers
•Reducing name server synchronization
•Reducing RSCN processing
The main reason for using AG mode is scalability. Given that embedded switches are smaller-port-count
switches, an environment with a lot of blade servers with embedded switches can easily start to encroach on
the stated limits on total domain count. Putting these switches into AG mode means they will not be consuming
domain. The downside to AG mode has been the functionality (or feature set) available, although AG continues
to expand its functionality with each release. Though there are some scenarios with a clear-cut answer for AG
mode, generally it is an evaluation of the SAN environment and the desired functionality that determines if AG is
a design option for the environment. In a fabric with lots of legacy devices, identifying and isolating misbehaving
devices is easier to do in a full-fabric environment.
Last, for congurations with hosts and targets on the same AG, the trafc does need to go through the fabric
switch, but it is handled within the local switch and does not need to traverse to another switch in the fabric and
then back again. The theoretical domain limit in a single fabric is 239, but most fabrics are typically limited to a