The primary objective of load testing is to simulate a realistic user workload to understand how a system will perform under expected, real-world conditions. Configuring TabJolt with a test plan that includes a diverse mix of dashboards, views, and user interactions (e.g., filtering, selecting marks) is essential for creating this realistic simulation. This approach ensures that the test results are representative of typical usage patterns, providing an accurate and effective assessment of Tableau Server's performance, capacity, and potential bottlenecks under normal operational load. Testing only the most intensive dashboards (stress testing) or simple ones would provide a skewed and incomplete picture of overall system behavior.

A. Installing TabJolt on the same machine as Tableau Server creates resource contention (CPU, memory), which contaminates the performance metrics and invalidates the test results.

C. This describes stress testing, a specific type of test to find a system's breaking point, not a comprehensive load test designed to model typical, accurate user behavior.

D. This is a logistical consideration to avoid disrupting production users but does not contribute to the intrinsic accuracy or effectiveness of the test workload itself.

1. Tableau Server Scalability: A Technical Deployment Guide for Server Administrators. (Official Vendor Documentation)

Reference: In the "Create a Test Plan" section

the guide states

"The most effective test plans are those that model the reality of your deployment... A good test plan will include a mix of workbooks that are representative of what your users will be doing on Tableau Server." This directly supports using a variety of dashboards representative of user behavior. The document also explicitly advises

"Do not install TabJolt on your production Tableau Server computers

" which invalidates option A. (See pages 10-11).

2. TabJolt Installation and User Guide. (Official Vendor Documentation - GitHub Repository)

Reference: The configuration of TabJolt

particularly through the vizpool.csv file

is designed to specify a list of views to be tested. The documentation implicitly and explicitly encourages populating this file with a set of workbooks that reflect the anticipated user workload to achieve a meaningful performance test. This supports the principle of testing a variety of representative actions.

3. Jain

R. (1991). The Art of Computer Systems Performance Analysis: Techniques for Experimental Design

Measurement

Simulation

and Modeling. Wiley. (Peer-reviewed Academic Publication/Textbook)

Reference: Chapter 3

"Workloads

" and Chapter 18

"Workload Characterization Techniques

" emphasize that a fundamental principle of any performance study is selecting a workload that is representative of the real system's usage. The text explains that using a non-representative workload is a common mistake that leads to inaccurate conclusions. This foundational academic principle directly supports option B. (DOI for a related paper by the author on the topic: https://doi.org/10.1109/2.58409)