Caching Heterogeneous Stuff in Heterogeneous Places
Session Recording and Materials
Join Red Hat Research for the next Research Days event, “Caching Heterogeneous Stuff in Heterogeneous Places” on August 10, 2022 from 3PM to 04:30PM IDT (2PM CEST, 8AM EDT).
Attend this talk for insight into some new algorithms that can help you make better use of your resources with dynamic workloads. During this talk Dr. Gabriel Scalosub and Dr. Gil Einziger of Ben Gurion University of the Negev, Israel, will discuss work that focuses on self-adjusting caching solutions, which show improved performance in various system environments. Ilya Kolchinsky, Research Supervisor, Red Hat, will lead the conversation. Attendees will have the opportunity to participate in the live discussion.
Caching is a classic optimization technique ubiquitous both within single hosts as well as in large distributed systems like CDNs, storage systems, databases, and data center networks. Classic solutions usually involve the omnipresent Least-Recently-Used (LRU) mechanism and its derivatives. These solutions are usually workload oblivious, and changing their configuration requires expert analysis of the system and the workload.
Recent advanced caching approaches that go well beyond LRU, which are gaining popularity, are self-adjusting and exhibit improved performance in various scenarios. We are harnessing some of the underlying design concepts to present novel caching frameworks that we are currently working on. These frameworks take into account request heterogeneity, and perform proactive caching decisions (a.k.a. speculative prefetching). We will be highlighting the design of algorithms and policies that are both dynamic and self-adjusting and consider their benefits in prominent storage system use cases.
Visit the Advanced Proactive Caching for Heterogeneous Storage Systems project page for more information on the project.
Ilya Kolchinsky, Research Supervisor, Red Hat