Ken Salem | 2009 Cheriton Research Symposium | SCS | UW

[Please remove <h1>]


Ken Salem – Client-Informed Caching for Storage Servers

Traditional caching policies are known to perform poorly for second-tier caches, such as those in storage servers.  I'll describe CLIC, a cache management scheme which can manage second-tier caches much more effectively. To achieve this, CLIC exploits hints that are generated by client applications.  CLIC is adaptive: it learns how to interpret the hints and use them to guide replacement in the second tier cache. I'll give some examples to illustrate hints and how they are used, and I'll show the results of a few experiments designed to measure CLIC's effectiveness.


Location: DC 1302
Time: 2:45-3:25pm