Class Central is learner-supported. When you buy through links on our site, we may earn an affiliate commission.

YouTube

One-Sided RDMA-Conscious Extendible Hashing for Disaggregated Memory

USENIX via YouTube

Overview

Save Big on Coursera Plus. 7,000+ courses at $160 off. Limited Time Only!
Explore a 15-minute conference talk from USENIX ATC '21 that delves into RACE hashing, an innovative one-sided RDMA-Conscious Extendible hashing index designed for disaggregated memory scenarios. Learn about the challenges of traditional hashing indexes in disaggregated memory environments and how RACE hashing addresses these issues through lock-free remote concurrency control and efficient remote resizing. Discover how this approach enables efficient indexing services using only one-sided RDMA verbs without involving compute resources in the memory pool. Examine the performance benefits of RACE hashing compared to state-of-the-art distributed in-memory hashing indexes, as demonstrated through extensive experimental results using YCSB hybrid workloads. Gain insights into the architectural overview, one-sided RDMA-Conscious subtables, and the Stale-read Client Directory (SRCD) concept.

Syllabus

Intro
Disaggregated Memory
Hashing Indexes in Local Memory
Hashing Indexes in Disaggregated Men
Hash Collision
Challenge 3: Extendible Resizing
Challenge Summary
RDMA-Conscious Extendible (RACE) H
Architectural Overview
One-sided RDMA-Conscious (RAC) Subtabl
Lock-free Remote Concurrency Control
Extendible Remote Resizing
Experimental Setup
Insertion
Update
YCSB Hybrid Workloads
The Stale-read Client Directory (SRCD)

Taught by

USENIX

Reviews

Start your review of One-Sided RDMA-Conscious Extendible Hashing for Disaggregated Memory

Never Stop Learning.

Get personalized course recommendations, track subjects and courses with reminders, and more.

Someone learning on their laptop while sitting on the floor.