Specifications
DATA CENTER BEST PRACTICES
SAN Design and Best Practices 24 of 84
•Brocade platforms are capable of generating SNMP traps for most error conditions. Consider implementing
some sort of alerting mechanism via SNMP.
Monitoring and Notications
Error logs should be looked at regularly. Many end users use combinations of syslog and SNMP with the
Brocade Fabric Watch and the logs to maintain a very close eye on the health of their fabrics. You can
troubleshoot network-related issues such as syslog events and SNMP traps through the Event Manager within
Brocade Network Advisor.
Brocade Network Advisor also collects, monitors and graphically displays real-time and historical performance
data, so you can proactively manage your SAN network.
Brocade Professional Services can be engaged to assist with implementing these and other advanced features.
Available Paths
It is recommended that the SAN be deployed with at least two paths between source and destination.
Often, there are more than two paths and the utilization of these paths is dependent on the routing
policy conguration.
•Port-Based Routing (PBR) assigns a single route between source port and destination port. Although
this minimizes disruption caused by changes in the fabric, it represents a less efcient use of
available bandwidth.
•Exchange-Based Routing (EBR) uses all available (equal-cost) routes between source port and destination
port, with individual exchanges assigned a single route. Although it represents a more efcient use of
available bandwidth, it is potentially more disruptive unless Dynamic Load Sharing (DLS) is implemented with
the lossless feature.
The number of available paths can be adjusted by changing the size of trunk groups. While a trunk can have
two to eight members, it may prove benecial to have more trunks with fewer members. Spreading ISLs across
multiple trunks uses more of the fabric bandwidth by spreading trafc across more paths. Keep at least two
members in each trunk to avoid unnecessary frame loss if a trunk member fails.
McDATA Interop Mode
A mixed fabric is one with Brocade FOS switches and McDATA Enterprise OS switches (M-EOS). The inter-
operability mode of the switch with McDATA switches can be McDATA Fabric mode, or McDATA Open Fabric
mode. (Refer to the Brocade FOS Release Notes for supported platforms and restrictions.) McDATA Open Fabric
mode is intended specically for adding Brocade FOS-based products to M-EOS fabrics that are already using
Open Fabric mode.
Brocade default routing protocol is EBR. McDATA default is PBR. Since these protocols operate very differently
in how trafc is distributed across the ISLs, the resulting lopsided ow control for ISLs causes performance
issues. McDATA’s “open trunking” can be used to assist with the reallocation of ows to better ISLs; however,
it will not be very effective against short duration ows (5–10 seconds) of small frames. This “micro-burst”
sustained over time can result in fabric-wide performance problems due to frame congestion. So, to mitigate
this possibility, or if you are experiencing this, you should try to identify these “heavy hitter” servers and when
(what time of day) the micro-bursts are happening. You will be able to correlate the performance spike with
errors on other non-related ows (collateral damage) from the non-discriminating C3 Discards on the ISLs they
are sharing. (From an M-Series [legacy McDATA] standpoint, a single ISL is used.)
Here are some recommendations, listed in order of their level of invasiveness. Some customers are willing to
move a server, but some are not. Also, some of these solutions will not work in certain environments, so begin
planning the appropriate solution as early as possible: