Check out the new USENIX Web site. next up previous
Next: Experimental Results Up: Experimental Set-up Previous: Compressing time


Comparison Metric

To evaluate the benefit from using various route control schemes we compare the response time of transfers when using the scheme (i.e., $Resp_{(x, scheme)}$, for a transfer $x$), with the response time when the best of the three providers is employed for each transfer ( $min_{i}\{Resp_{(x, ISP_i)}\}$):


$\displaystyle \mathcal{R}_{scheme}$ $\textstyle =$ $\displaystyle \frac{1}{\left\vert x\right\vert}\sum_{x}\frac{Resp_{(x, scheme)}}{min_{i}\{Resp_{(x, ISP_i)}\}}$ (1)

Where, $\left\vert x\right\vert$ is the total number of transfers. We call $\mathcal{R}$ the ``performance metric'' or the ``normalized response time''. The closer $\mathcal{R}$ is to 1, the better the performance of the scheme.

In the above computation, the response times from employing the best provider for any transfer (the terms in the denominator above) are computed in an offline manner for each transfer by forcing it to use each of the three providers and selecting the provider offering the best response time.



Anees Shaikh 2004-05-05