IBM Tivoli Monitoring: Implementation and Performance Optimization for Large Scale Environments

112 IBM Tivoli Monitoring: Implementation and Performance Optimization for Large Scale Environments

groups with a collection interval of 5 minutes (1,1 MByte per agent (see the

example in 4.3.2, “Estimate agents disk space for data collection using the

Warehouse Load Projections spreadsheet” on page 94), then the amount of

collected will not be more than 5 GB per day.

If the Warehouse Proxy agent has a 100 Mb/s interface, then less than 1% of the

bandwidth will be used. But the same consideration holds true as discussed in

4.3, “Size the Tivoli historical data collection” on page 80. That is, if there are

attribute groups with multiple instances, then the value can be multiples of this

number. In every case, the initial value is the “Bytes per instance (agent)” and

this amount of data in most cases goes in and out the Warehouse Proxy agent.

As shown in Figure 4-22 on page 109, you can put the remote monitoring server

and the associated Warehouse Proxy agents on one system. The memory and

disk requirements have to be added and you should follow the recommendation

to connect not more than 1000 agents to the remote monitoring server and so to

the Warehouse Proxy agent, as well. The bandwidth for the Warehouse Proxy

agent should be estimated roughly.

4.5.2 Determine the upload time

Another question can be how to specify the time that data will be uploaded from

the Warehouse database via the Warehouse Proxy agent, particularly when the

Warehouse interval is configured to run once a day. There are three

considerations in determining this time:

򐂰 Adjusting the upload time to run outside the office hours for saving bandwidth

򐂰 Synchronizing the upload time of detailed data with the Summarization and

Pruning agent

򐂰 Synchronizing the upload time with creation of a data mart (if used)

There is no command line interface to configure or start the historical data

collection. Only the Historical Data Collection dialog box can be used for start the

collection.

If you start a data collection for an attribute group, the data will be written on the

local disk. Approximately one hour or one day later (depending on the chosen

Warehouse interval), the upload process will start. That means, for example, if

there is a Warehouse interval of one day and the collection will be started at 4

a.m., then the first upload will be done at 4 a.m. the next day.

If there are a large number of agents for the selected attribute group, it will take a

long time to upload all data from all agents into the Warehouse database. But

there is no guarantee that the upload interval for the agents can be adjusted in

the early morning time, as in this example. Downtimes of agents, monitoring