Advantages of Simulation Modeling

Our ability to what-if is limited by the constraints of time and information. It takes time to understand a complex environment and all its subtle interactions. There is the time expended in staying current with the constant changes in a dynamic services ecosystem. It takes additional time to work through a what-if scenario. The iterative nature of a what-if exploration really compounds the time constraint. The usual process is seeing a change that indicates an improvement, problem cause, or a sensitivity factor and then pursuing it further.

The next questions that naturally arise are as follows:

• Is this change really causing the improvement?

• Where are we in the variable's range? Will a bigger change to it lead to a bigger improvement?

• Are there any instabilities we should know about, such as scaling under loads?

• Are there simpler or cheaper alternatives?

Because of limited time and resources, some organizations trying to answer these questions often limit an approach to what has been done in the past. They are playing it safe, but missing opportunities to make a bigger impact. The other risk is missing key trends by focusing on the familiar. More than a few decisions have also been based upon just plain guessing and hoping for the best.

Simulation modeling quickly explores many options, leading to better understanding and decision making. The benefits of simulation modeling are as follows:

• Easy exploration of a wide set of alternative approaches

• Independence from a test bed

• Flexibility

After a simulation model is constructed and validated, it can be used for exploring a range of alternatives for planners, administrators, and designers. They can quickly eliminate those alternatives that do not improve performance or service quality. Rapidly iterating through alternatives leads to an optimum solution for a set of operating scenarios. Being able to evaluate alternative designs and workloads without modifying hardware and software can be a definite advantage.

Evaluating a range of alternatives is also helpful in identifying sensitivities. For example, changing the loading characteristics, the transaction mix, the topology, or other factors can identify specific sensitivities. A certain mixture of services may introduce instability and mutual interference, while the same mixture with different proportions operates smoothly.

Simulation modeling tools allow more agility and faster results when compared to using load testing on a test bed (which is discussed in Chapter 11). For example, with a model, you can add a different kind of device or one that is needed and not yet delivered to the test bed. This enables testing and analysis to go forward without waiting for all the real pieces of the environment to be assembled. The elements of a model can be updated, replaced, or modified in a matter of minutes and new results can be produced quickly thereafter. In contrast, ordering all possible products that can be placed in the test bed is not economically feasible. Even if it were, delays in obtaining the products and integrating them into the test bed must be considered. Modeling is especially useful when parts of the physical or software infrastructure are not available and testing can begin without them.

Results are usually obtained faster with modeling than with load testing on a test bed because there is no physical infrastructure to deal with and no software to modify. For example, making a change to the test bed requires staff time to create any physical changes such as altering connectivity, reassigning servers, moving switches, or changing link capacity. Additional time may be required to modify software, update directories, and adjust management tools to reflect changes. Further effort is needed to verify that the changes were made properly and introduced no new sources of errors or problems.

The expenditures for the equipment in the test bed, for its administration, and for its operation must also be considered relative to the costs for acquiring and learning to use good modeling tools. It's a matter of balancing your investment strategy and making sure that the complete suite of tools provides the most cost-effective management capabilities.

Some organizations have their integrator maintain a model that they use for planning and what-if scenarios. Others aquire load testing tools or work with testing organizations.