This page provides room for information about the agent, that is important or could be useful for this project.
The information on this here can provide a short overview, but is not intended as user documentation about InspectIT.

...

Framework

The agent uses the PICO Container PicoContainer to handle class instantiation. The container is following the principle of Dependency Injection and Inversion of Control.

...

• Reading and applying of the provided configuration files
• Gathering of invocation data
• Communication with the server . (Sensor registration etc.)
• Temporal Storage storage of invocation data , before its send to CMR
• Performing the sensor logic
• Hook Injectioninjection

Configuration

Each agent has a configuration that provides IP-address and port, via which he can communicate with the CMR and send data to.
Also the sensors that have to be used and the classes that have to be instrumented are configured in the configuration file.
Also a buffer strategy has to be set, so that the agent knows, what to do when it's not possible to send the data at the moment.
Its important to keep in mind, that a high aggregation of the collected data has an negative impact on the CPU consumption of the target application, bus puts less stress on the network.

JVM

...

environment

Class Loading

The classes are loaded at that moment, when there are needed for the first time. Not earlier.
Static classes and that containing static methods are instantiated on application startup and are not effected by this the upper rule.
When a class is needed, the class loader loads the class and then it's linked:.

There are three phases in the linking process:

...

There is no need to instrument classes with the class loader method anymore.

For bytecode modification the class library Javassist is used.

Threads

The system thread is created, when the method start() in the thread object is called. That means, that there is no concurrency involved, since the run() method is only called by the new system thread.

Three points of interest are essential on threads:

• Which thread created a the current thread
• Which one starts it
• Which thread executes it

...

Its important to keep the principle of thread pools in mind. Its commonly used by web applications.
It allows the reuse of threads, which means there is no need to create new thread instances. This reduces overhead, since the system don't have to manage new threads(no new memory allocation etc.). .

Runnables

Runnables can be also reused.

But a thread cannot switch the Runnable during execution.