Check out the new USENIX Web site. next up previous
Next: Achieving Exclusivity Up: On Multi-level Exclusive Caching: Previous: Bounds for Multi-path Hierarchies


The PROMOTE Technique

Figure 4: The common methods used by PROMOTE to adapt $ probPromote$ within every cache by leveraging the hint protocol.
\begin{figure}{\small
\vspace*{2mm}\hrule\vspace*{1mm} {\small {\raggedright {\s...
...lse;$\\
21:\> endif\\
22:\> return $true;$
\end{tabbing}\hrule
}\end{figure}

Figure 5: The PROMOTE augmentations to ARC and LRU algorithms. These enhancements are apart from the regular mechanisms (not shown) inherent in these algorithms.
\begin{figure}{\small
\hrule\vspace*{1mm} {\small {\raggedright {{\sc Promote-LR...
...igher cache ($p$, $promoteHint$, $T_2hint$)
\end{tabbing}\hrule
}
\end{figure}

The goal of the PROMOTE technique is to provide exclusive caching that performs better than DEMOTE, while at the same time, requires no demotions, reloads from disks, or any computationally intense operation. Each cache independently and probabilistically decides whether to keep the data exclusively as it is passed along to the application. The probability of holding the data or promoting the data upwards is adaptively determined. We present the PROMOTE algorithm in Figures 4 and 5.



Subsections

root 2008-01-08