Главная
Study mode:
on
1
Intro
2
Data replication: when server crashes...
3
Raft: distributed consensus
4
Erasure Coding: Reed-Solomon Code
5
Combine erasure coding with crash tolerance?
6
CRaft: erasure coding + Raft
7
Challenge #1: liveness
8
Challenge #2: newly elected leader
9
Comparing consensus protocols
10
Evaluation highlights
11
Evaluation setup
12
Evaluation: read latency
13
Summary: Craft
14
Evaluation: liveness
Description:
Explore a 27-minute conference talk from USENIX FAST '20 that introduces CRaft, an innovative version of the Raft consensus protocol incorporating erasure coding to reduce storage and network costs. Delve into the challenges of combining erasure coding with crash tolerance in distributed systems, and learn how CRaft maintains the same liveness as Raft while achieving significant improvements in storage efficiency, write throughput, and latency. Examine the implementation of CRaft in a key-value store and analyze its performance compared to the original Raft protocol. Gain insights into the technical aspects of consensus protocols, data replication strategies, and the application of Reed-Solomon codes in distributed systems.

CRaft - An Erasure-coding-supported Version of Raft for Reducing Storage Cost

USENIX
Add to list
0:00 / 0:00