Popularity Prediction for Social Media over Arbitrary Time Horizons
Daniel Haimovich, Dima Karamshuk, Thomas Leeper, Evgeniy Riabenko, Milan Vojnovic
USENIX Annual Technical Conference (ATC)
Server Flash Cache (SFC) is being increasingly adopted in virtualization environments for IO acceleration. Deciding the optimal SFC allocation among VMs or VM disks is a major pain-point, dominantly handled manually by administrators. In this paper, we present vCache- Share, a dynamic, workload-aware, policy-driven framework for continuous and automated optimization of SFC space partitioning. Its decision-making is based on multiple IO access characteristics. In particular, SFC adopts a novel cache utility model that captures both longerterm locality behavior and transient locality spikes. This paper validates the growing applicability of analytical programming techniques to solve real-time resource management problems, traditionally addressed using heuristics. We designed vCacheShare to coordinate with typical VMmobility events and implemented it within the widely used ESXi hypervisor. We performed extensive evaluation using 13 representative enterprise IO workloads, one IO benchmark, and two end-to-end deployment test cases targeting Virtual Desktop Infrastructure (VDI) and data warehousing scenarios respectively. Our results verified the advantage of vCacheShare over implicit management schemes such as global LRU, and confirmed its self-adaptive capability.
Daniel Haimovich, Dima Karamshuk, Thomas Leeper, Evgeniy Riabenko, Milan Vojnovic
Liqi Yan, Qifan Wang, Yiming Cu, Fuli Feng, Xiaojun Quan, Xiangyu Zhang, Dongfang Liu
Patrick Lewis, Barlas Oğuz, Wenhan Xiong, Fabio Petroni, Wen-tau Yih, Sebastian Riedel