Adaptive Online Cache Capacity Optimization via \nLightweight Working Set Size Estimation at Scale
Year: 2023
Type: conference-paper
Abstract: Big data applications extensively use cache techniques to accelerate data access. A key challenge for improving cache \nutilization is provisioning a suitable cache size to fit the dynamic working set size (WSS) and understanding the related \nitem repetition ratio (IRR) of the trace. We propose Cuki, an \napproximate data structure for efficiently estimating online \nWSS and IRR for variable-size item access with proven accuracy guarantee. Our solution is cache-friendly, thread-safe, \nand light-weighted in design. Based on that, we design an \nadaptive online cache capacity tuning mechanism. Moreover, \nCuki can also be adapted to accurately estimate the cache miss \nratio curve (MRC) online. We built Cuki as a lightweight plugin of the widely-used distributed file caching system Alluxio. \nEvaluation results show that Cuki has higher accuracy than \nfour state-of-the-art algorithms by over an order of magnitude and with better stability in performance. The end-to-end \ndata access experiments show that the adaptive cache tuning \nframework using Cuki reduces the table querying latency by \n79% and improves the file reading throughput by 29% on average. Compared with the cutting-edge MRC approach, Cuki \nuses less memory and improves accuracy by around 73% on \naverage. Cuki is deployed on one of the world’s largest social \nplatforms to run the Presto query workloads.
Authors: Gu, Rong, Li, Simian, Dai, Haipeng, Wang, Hancheng, Luo, Yili, Fan, Bin, Basat, Ran Ben, Wang, Ke, Song, Zhenyu, Chen, Shouwei, Wang, Beinan, Huang, Yihua, Chen, Guihai
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Keyword: online estimation
Open Access status: green
OpenAlex ID: W7026541320
