10.7 Summary

---

10.7 Summary

In this chapter we introduced many new concepts. The first was the use of testbed systems to allow the modeler to construct subsets or even entire systems to be used as environments in which to iron out the performance issues with new or existing designs before they are built or before an existing system is altered. This presentation was then followed by an example of a network testbed used to analyze protocols. This example helped the reader to understand the ways in which a testbed can be effectively used as part of a larger modeling effort.

This discussion was then followed by a general discussion of measurement techniques applied within testbed and existing systems. The primary techniques described are hardware monitors and software monitors. The benefits and shortfalls of each technique were discussed along with examples of how they may be applied within a modeling project.

The discussion then changes to looking at the development of workloads to drive testbed environments. We focus on the types of workloads and how the modeler decides on which level of detail to focus the workload, how to determine the system and components the workload is to stress, and what types of services are to be mimicked. This is then followed by a presentation dealing with structuring the workload as either an average load, a real load, or something in between. We follow this up with a discussion of testbed result presentation and some concepts concerning good practice in presenting performance study outputs to the client.

---