A baseline is defined as the accepted attributes that describe a system at a particular point in time. Thus, the baseline serves as a point of reference. The idea is to capture performance metrics after every change and determine their effectiveness by comparing them to the baseline. Changes can only be compared one at a time. While working with baselines, we need to make sure that all aspects except for the single change must remain the same. Thus, the new metrics data, after the change, when compared to the baseline data, will show whether the performance improves or declines.
Our goal at the end of the chapter is to generate baselines that can be used to find out the efficacy of optimizations.
Performance tests executed on various tools cannot be compared directly with each other. The tools vary significantly in how they simulate load and how metrics are determined; thus, the results vary largely when compared with each other. But the trend generated by tests on one tool is comparable to a trend from another tool. If a tool demonstrates a decrease in performance, then the other tool should provide a similar result.
Load simulation opens sockets on the client side. The sockets, in turn, are treated as file descriptors. Thus, make sure there are enough file descriptors available on the box. The limit can be enhanced by the ulimit -n command or by changing the security.limits file.