Suggestions on Testing of Related Factors

I am designing an experiment that measures the performance of a computer-based system as the response. The response will be measured by how long it takes users of the system to perform a task using the system, under certain conditions. One of the things the system does is receive messages that describe entities, compare the message contents to entities that reside in the system, and decide whether to reject duplicate data, create a new entity, or update an existing entity.

Two of the factors that may affect the system performance are: Message Rate (messages received per unit time) and Number of Entities (i.e., entities residing in the system, to which new messages are compared). I can control the Message Rate, the content of the messages (e.g., updates vs. new entities), and the initial Number of Entities.

The confusing part is that the Number of Entities will not remain constant, as incoming messages result in new or updated entities. It bothers me that the Message Rate (combined with the message content) and the Number of Entities are related, and also that I cannot keep the Number of Entities constant over the course of a single test run.

Any suggestions would be appreciated.
The number of entities sounds like a queue. Arrivals in the queue are based on the (presumably poisson) message rate. entities leave the queue according to what the system does with them.....or maybe I misundestood ?
queue comments

There is an input queue, not explicitly what I am concerned about. They become entities resident in the system.

When the system takes a message from the input queue, it processes them through the algorithm that determines whether they are new entities or updates to existing entities held in the system. It does this by comparing them to the existing entities it holds. The system will take more or less time, depending on whether they are new or updates.

Thanks, Alan
Thank you for the question. . . . I realize now that the real factors on system performance are the rates, which I can divide into a rate of new entity reports and a rate of entity updates.