Jeffrey K. MacKie-Mason

The Case for Market-based Push Caching (Download full paper)

MacKie-Mason, Jeffrey K. Chan, Yee Man Womer, Jonathan Jamin, Sugih

Published on: November, 1999

Variable QoS from Shared Web Caches: User-Centered Design and Value-Sensitive Replacement (Download full paper)

Terence P. Kelly, Sugih Jamin and Jeffrey K. MacKie-Mason

Published on: June, 1999

Abstract: Due to differences in server capacity, external bandwidth and client demand, some Web servers value cache hits more than others. Assuming that a shared cache knows the extent to which different servers value hits, it may employ a value-sensitive replacement policy in order to generate maximum aggregate value for servers. we consider both the prediction and value aspects of this problem and introduce a novel value-sensitive LFU/LRU hybrid which biases the allocation of cache space toward documents whose origin servers value caching most highly. We compare our algorithm with others from the Web caching literature and discuss from an economic standpoint the problems associated with obtaining servers' private valuation information.

Biased Replacement Policies for Web Caches: Differential Quality-of-Service and Aggregate User Value (Download full paper)

MacKie-Mason, Jeffrey K. Kelly, Terence P. Chan, Yee Man Jamin, Sugih

Published on: January, 1999

Abstract: Disk space in shared Web caches can be diverted to serve some system users at the expense of others. Cache hits reduce server loads, and if servers desire load reduction to different degrees, a replacement policy which prioritizes cache space across servers can provide differential quality-of-service (QoS). We present a simple generalization of least-frequently-used (LFU) replacement that is sensitive to varying levels of server valuation for cache hits. Through trace-driven simulation we show that under a particular assumption about server valuations our algorithm delivers a reasonable QoS relationship: higher byte hit rates for servers that value hits more. We furthermore adopt the economic perspective that value received by system users is a more appropriate performance metric than hit rate or byte hit rate, and demonstrate that our algorithm delivers higher "social welfare" (aggregate value to servers) than LRU or LFU.

One Size Doesn't Fit All: Improving Network QoS Through Preference-driven Web Caching (Download full paper)

Chan, Yee Man, Jonathan Womer, Jeffrey K. MacKie-Mason and Sugih Jamin

Abstract: In order to combat Internet congestion Web caches use replacement policies that attempt to keep the objects in a cache that are most likely to get requested in the future. We adopt the economic perspective that the objects with the greatest value to the users should be in a cache. Using trace driven simulations we implement an incentive compatible market-based Web cache for servers to push content into a cache. This system decentralizes the caching process as servers provide information in the form of bids for space in the cache. Truthful information from the server on valuations of objects and predictions of hit rates is obtained. This information is used in filling the cache, which can provide increased aggregate value and differential quality of service to servers when compared to LFU and LRU.