Академический Документы
Профессиональный Документы
Культура Документы
com
Agenda
Introduction Object Server Key Performance Indicators Probe and Gateway Key Performance Indicators Q&A
Introduction
To check that Omnibus is performing well, there are several key performance indicators that can be monitored. When using the key performance indicators, first establish a baseline on the system when it is under normal load and operation. The Key Performance indicators can be used to measure performance when changes are made to the environment, by comparing the baseline to the KPI measurements after the change.
Mon Oct 12 18:03:56 2009: Trigger Group 'gateway_triggers' Mon Oct 12 18:03:56 2009: Time for all triggers in report period (60s): 29.789663s
10
High generic_clear or deduplication triggers can indicate high event throughput or high number of resident events. Ensure best practices are used in creating custom triggers http://publib.boulder.ibm.com/infocenter/tivihelp/v8r1/topic/com.ibm.netcool_OMNIbus.doc _7.3.1/omnibus/wip/admin/reference/omn_adm_per_bestpracticestriggers.html Ensure trigger execution time is kept to a minimum, no other writes can be performed in the Object Server when a trigger is executed.
11
Number of Rows
alerts.status and alerts.journal alerts.details alerts.details table should only be used when alerts.status is not enough to hold enough information for a specific alarm or during rules file development. On production systems, it is suggested to keep the alerts.details table below 5,000 rows. If you have a large number of rows in alerts.details, the ObjectServer performance can be degraded. Details statement in probe rules file are used to generate records into alerts.details table. details($*) will record each token as one row into alerts.details. If you have details($*) enabled in your rules file, for each event in alerts.status table, you might have 10~50 rows in alerts.details table. Details can be disabled by commenting out any details($*) statements in all your probe rules file, restarting all probes, clearing the current records in details table (with "delete from alerts.details")
12
13
15
Stats triggers
The data gathered by this trigger group and automation is written periodically to the master.stats table. The default write interval is 300 seconds; this value is configurable in the statistics_gather trigger 5 days of data is retained by default
16
CPU usage
Monitor the CPU usage of the nco_objserv process If the Object Server is under heavy load, this will be reflected in CPU usage Profiler report and trigger statistics logs will show the source of the heavy load Sizing considerations https://www.ibm.com/developerworks/mydeveloperworks/wikis/home? lang=en#/wiki/Tivoli%20Netcool%20OMNIbus/page/OMNIbus%20Sizing %20Guide
17
Memory Usage
Memory usage of nco_objserv process The memory usage of the process increases proportionally to increases in the number of rows in the alerts.status table, alerts.details table, and the alerts.journal table (or any additional tables you have defined), to increases in the number of connections, and increased usage by clients. The memory usage should remain stable over time, and any increases should correspond to increases in the numbers of table rows or additional clients. Sizing considerations https://www.ibm.com/developerworks/mydeveloperworks/wikis/home? lang=en#/wiki/Tivoli%20Netcool%20OMNIbus/page/OMNIbus%20Sizing %20Guide
18
19
20
21
22
23
24
Summary
Object Server Key Performance Indicators Object Server Granularity Profiler Report Trigger Stats Number of rows in alerts.status, alerts.journal and alerts.details Number of inserts in the alerts.status table Number of Connections CPU usage of nco_objserv Memory Usage of nco_objserv Memstore usage Probe Key Performance Indicators Number of events received CPU Usage Memory Usage Average Time spent processing rules Gateway Key Performance Indicators CPU Usage Memory Usage
25 2010 IBM Corporation
Questions?
27