Specifications
DATA CENTER BEST PRACTICES
SAN Design and Best Practices 22 of 84
Bottleneck detection can also serve as a conrmation to host information when storage latencies are suspected
in poor host performance. The reverse (eliminating the storage as the source of poor performance) is also true.
Beginning with Brocade FOS v6.4, Bottleneck Detection can also be applied to ISLs (E_Ports) and will highlight
issues on those links.
The sampling interval and number of notications are congurable, as well as the alerting mechanisms. With
Brocade FOS v6.4 notications can be congured for Reliability, Availability, and Serviceability (RAS) log and
Simple Network Management Protocol (SNMP). Brocade Network Advisor can be congured to automatically
monitor and detect bottlenecks in the fabric. You can easily pinpoint areas of network congestion with visual
connectivity maps and product trees.
Design Guidelines
Edge Hold Time (EHT):
•Recommended primarily for initiators (hosts). Extreme care must be taken if you choose to apply EHT to target
ports because a target port can service a large number of initiators. A large number of frame drops on a
target port can potentially affect a very large number of running applications. Those applications may be more
tolerant to poor performance that to a large number of I/O retries.
•There is no calculation for determining the best value for EHT. EHT can be set from 100 to 500 milliseconds.
The lower the value, the more frame drops you can expect. Brocade recommends that you take a value of
approximately 250 milliseconds and observe the results.
•EHT is less effective when initiators and targets share the same switch, because the timeout value will apply
equally to both storage and host ports.
•EHT applies to the entire ASIC. If possible, ISLs should be placed on a different ASIC than the servers.
Bottleneck Detection:
•A phased approach to deploying Bottleneck Detection works best. Given the potential for a large number
of alerts early on in the process, Brocade recommends starting with a limited number of storage ports and
incrementally increasing the number of ports monitored over time. Once the storage latencies are dealt with,
you should move on to the host (initiator) ports and ISLs. You can increase the number of ports monitored
once the chronic latency problems have been dealt with.
•Bottleneck Detection consumes some switch memory to keep some historical data. Brocade recommends no
more than 100 ports in total be monitored at once on Brocade 48000 platforms, to avoid any potential for
memory issues. There are no such limitations on the Brocade DCX 8510 with Gen 5 Fibre Channel or
DCX platforms.
AVAILABILITY AND HEALTH MONITORING
With Brocade FOS and Brocade Network Advisor, IT organizations can monitor fabrics on both a real-time
and historical basis. This allows users to address performance issues proactively and rapidly diagnose
the underlying causes, then quickly resolve the issues before the SAN becomes the bottleneck for critical
applications. An overview of the major components is provided below. A complete guide to health monitoring
is beyond the scope of this document. Please refer to the Brocade Fabric OS Command Reference Guide, the
Brocade Fabric OS Troubleshooting Guide, the appropriate Brocade SAN Health and Fabric Watch guides, and the
Brocade Network Advisor SAN User Manual for more detailed information.
Health Monitoring
Brocade Fabric Watch
Fabric Watch is an optional health monitor that allows you to constantly monitor each director or switch for
potential faults and automatically alerts you to problems long before they become costly failures.