Fabric OS Administrator's Guide v7.0.0 (53-1002148-02, June 2011)

Fabric OS Administrator’s Guide 301

53-1002148-02

Bottleneck detection overview

• How many affected seconds are needed to generate the alert.

• How long to stay quiet after an alert

Changing alerting parameters affects RASlog alerting as well as SNMP traps.

Using alerting parameters to determine whether alerts are generated

You have the option of receiving per-port alerts based on the latency and congestion history of the

port. Alerts are generated based on the number of affected seconds over a specified period of

time. If the number of affected seconds is higher than the threshold, an alert is generated. This

process is done independently for latency and congestion.

The bottleneckmon alerting parameters determine whether an alert is generated.

For example, Figure 53 shows an interval of 12 seconds, in which 6 seconds are affected by a

congestion bottleneck and 3 seconds are affected by a latency bottleneck.

FIGURE 53 Affected seconds for bottleneck detection

The -time parameter specifies the time window. For this example, -time = 12 seconds.

The -cthresh and -lthresh parameters specify the thresholds on number of affected seconds that

trigger alerts for congestion and latency bottlenecks, respectively. For this example, assume the

default values for these parameters:

• -cthresh = 0.8 (80%)

• -lthresh = 0.1 (10%)

For this time window, 50% of the seconds (6 out of 12 seconds) are affected by congestion. This is

below the threshold of 80%, so an alert would not be generated for a congestion bottleneck.

For the same time window, 25% of the seconds (3 out of 12 seconds) are affected by latency. This

exceeds the threshold of 10%, so an alert would be generated for a latency bottleneck.

Time (seconds)

Affected by congestion bottleneck?

Yes

0123456789101112

Time (seconds)

Affected by latency bottleneck?

0123456789101112