HP StorageWorks Fabric Manager 5.x Administrator Guide (AA-RUQSF-TE, September 2005)

88 Using the Performance Monitor
End-to-end monitoring
End-to-end monitoring sets all possible target and initiator pairs from the filtered list retrieved from the API
for the fabric. Fabric Manager stores the configuration of the monitors persistently and periodically
re-creates them on the fabric as necessary. Each time the Fabric Manager server retrieves the monitor
values, it checks for and re-creates any monitors (configured through Fabric Manager) that are no longer
present. If a resource limit is reached while trying to set all of the possible pairs, performance monitoring
does not attempt to determine the best set of monitors. Instead, monitors are set on a first-come, first-serve
basis (in the order presented to Fabric Manager from the API).
Configuration of data-capture file information is not supported. The monitor values for the end-to-end
monitors are retrieved at a set interval (which is not configurable) and stored in the database from the
previous release. You can then request that reports be generated from the stored data (see ”Generating
custom reports and graphs” on page 92).
Fabric Manager attempts to set the monitors between the following types of devices only:
• Initiator-target
• Initiator-unknown
• Target-unknown
• Initiator and target-any device role
• Unknown-any device role
Monitors between initiator-initiator and target-target are not set. Monitors between all other device types
are valid.
Granularity
Granularity is the timeframe for the sample values that are used to generate a report. Since Fabric
Manager stores a limited number of samples for each granularity, every level of granularity is not
available for all possible time ranges. For example, if you select a timeframe to be the past three days, the
five-minute granularity level is not available. The options for granularity (for both Port Statistics and
End-to-End Monitoring) are listed in Table 14.
There might be gaps in the stored data. Events that can cause gaps in data include:
• Authentication failures: Fabric Manager uses switch user names and passwords to store data. If the
password is changed (resulting in an authentication failure) on a switch where performance
monitoring is enabled, data collection fails (creating a gap in the stored data).
• Switch not reachable: Data is not gathered if the launch switch is unreachable. If the launch switch is
reachable, but additional switches in the fabric are unreachable by Ethernet, the data is gathered.
• Performance monitoring is not enabled: Data is not being gathered, causing a gap in the stored data.
NOTE: If data is not available for some of the granularity points when values are rolled up to the next
time period value, the missing data is ignored. For example, if five of six 5-minute values are 50, and the
sixth value is missing: N1= 50, N2= 50, N3=50, N4= 50, N5= 50, N6= NA. The rolled up value for
30-minute entry=250 (because the N6 value is calculated as if it is a value of zero).
Table 14 Stored samples (per fabric) for granularity options
Granularity Samples Time covered in stored samples
5 minutes 600 2 days and 2 hours
30 minutes 700 2 days and 2 hours, plus 12.5 days
2 hours 775 2 days and 2 hours, plus 12.5 days, plus 50 days
1 day 797 2 days and 2 hours, plus 12.5 days, plus 50 days, plus 732 days
(rounded to 797 days)