CSD Researchers Earn Community Award for Cache-Efficiency Research

Rashmi Vinayak, Juncheng Yang Among Paper's Authors

Wednesday, May 8, 2024 - by Marylee Williams

The Computer Science Department's Rashmi Vinayak and Juncheng Yang were among the authors of a paper on cache-eviction algorithms that won the Community Award at the 2024 USENIX Symposium on Networked Systems Design and Implementation.

Researchers from Carnegie Mellon University's Computer Science Department won the Community Award at the 2024 USENIX Symposium on Networked Systems Design and Implementation (NSDI) for work on a new cache-eviction algorithm.

The Community Award is given to the best paper where the code, dataset or a combination of both are made publicly available.

The paper, "SIEVE Is Simpler Than LRU: An Efficient Turn-Key Eviction Algorithm for Web Caches," details a cache-eviction algorithm that achieves state-of-the-art efficiency while maintaining simplicity for web cache workloads.

Caching is critical for ensuring webpages load quickly. It's like a memory bank, storing items such as images or icons that need to be accessed quickly. A cache doesn't have unlimited space, though, and has to be cleared, which is where the eviction algorithm is used. While least recently used (LRU) and first in first out (FIFO) are the two most common eviction algorithms, they aren't always the most effective. Over the years, people have generated huge amounts of data, little of which is accessed often or quickly. For example, when you browse your friend's photos from 10 years ago, the photos are loaded into the cache for quick access. But other people probably won't access them, so the photos take up precious space before being evicted.

SIEVE observes that these photos probably won't be accessed soon and uses this knowledge to quickly evict the photos to make room in the cache for more useful data. Engineers from different companies have already implemented over 20 open-source SIEVE cache libraries in more than 12 programming languages.

The paper's authors include Juncheng Yang and Rashmi Vinayak in CMU's Computer Science Department; Yazhuo Zhang from Emory University; Ymir Vigfusson from Emory University and Keystrike; and Yao Yue from the Pelikan Foundation.

Learn more about the team's research on SIEVE's website.

For more information, Contact:
Aaron Aupperlee | 412-268-9068 | aaupperlee@cmu.edu