About job throttling

Job throttling enables the ESM manager to distribute policy run jobs to multiple agents at a time. The number of agents is a configurable value on manager, which you can specify in the manager.conf file. The manager.conf file is located in the config folder that exists in the directory where Symantec Enterprise Security Manager is installed.

When you start a policy run on a domain, the Symantec Enterprise Security Manager manager checks the following parameters in the manager.conf file:

AGENT_STARTERS

Specifies the number of agents on which a manager initiates simultaneous policy runs.

The value can be between 1 to 5 for both Windows and UNIX managers. The default value for both Windows and UNIX managers is 1.

The recommended value that you should specify is 5.

AGENT_STARTER_TIMEOUT_SECONDS

Specifies the duration for which the Symantec Enterprise Security Manager manager waits to get a response from the agent before it initiates the policy run.

The default value is 360.

AGENT_STARTER_LIVEUPDATE_TIMEOUT_SECONDS

Specifies the duration for which the Symantec Enterprise Security Manager manager waits to get a response from the agent that can be updated using LiveUpdate.

The default value is 600.

The ESM manager reads and uses these parameters as follows:

Simultaneously executes the policy run request on the number of agents that is specified in AGENT_STARTERS.
Waits for the time that is specified in AGENT_STARTER_TIMEOUT_SECONDS to get a response from any of the agents.
Executes the policy run request on the next agent after either a timeout or after the manager receives a response from any agent.
Repeats this process until the manager contacts the specified agents to initiate the policy run.

Job throttling shows considerable improvement in the time that is required to complete policy runs when a few agents are not contactable. We have observed that in a customer environment, usually 30% agents remain unreachable. The improvement is based on the comparison between policy run time, where AGENT_STARTER = 1 and AGENT_STARTER = 5.

In our test, we have seen an improvement of 77% for Windows managers and 70% on Solaris managers for policy runs.

You may observer a lower or higher rate of improvement depending on the following parameters:

The number of unreachable agents	More number of unreachable agents results in a higher percentage improvement.
The rate of frequency when the manager finds unreachable agents	The percentage improvement shows better results if the unreachable agents are clustered together.
The time that the contactable agents take to complete the policy run	The time that the agents take to complete a policy run inversely impacts the percentage improvement.

If all the agents in your network are contactable by the manager, then the percentage improvement may be low.