Mean response time

In this section, we study the mean response time of the beneficiary jobs and the donor jobs, analyzed via DR. Throughout we will use the term ``gain'' to denote the improvement (drop) in mean response time experienced by beneficiary jobs under cycle stealing, as compared with dedicated servers (without cycle stealing), and the term ``pain'' to refer to the increase in mean response time experienced by donor jobs under cycle stealing as compared with dedicated servers:

$$\mbox{gain} = \frac{\mbox{{\bf\sf E}}\left[ T_B \right]^{{\t... ... CS}}}{\mbox{{\bf\sf E}}\left[ T_D \right]^{{\tt Dedicated}}},$$

where $\mbox{{\bf\sf E}}\left[ T_B \right]^{{\tt Dedicated}}$ refers to the mean response time of beneficiary jobs under dedicated servers and $\mbox{{\bf\sf E}}\left[ T_B \right]^{{\tt CS}}$ refers to the mean response time of beneficiary jobs under cycle stealing. $\mbox{{\bf\sf E}}\left[ T_D \right]^{{\tt Dedicated}}$ and $\mbox{{\bf\sf E}}\left[ T_D \right]^{{\tt CS}}$ are defined similarly. Observe that both pain and gain have been defined to range from $1$ to $\infty$, where infinite gain corresponds to the situation where the mean response time under dedicated is infinite while it is finite under cycle stealing. In Figures 6.4-6.7, we fix the threshold values as $N_B^{th}=N_D^{th}=1$ and study the effect of other parameters, and in Figures 6.8-6.9 we study the effect of the threshold values.