Distributed Systems 6.1: Consensus

preview_player
Показать описание
Рекомендации по теме
Комментарии
Автор

I accidently came to this channel and noticed that you are the famous Martin Kleppmann, author of the 'Designing Data Intensive Applications'. Nice to meet you Martin, I really enjoyed reading your book.

daratha
Автор

Fault-tolerant total order broadcast [0:15]
Manual failover [1:15]
Consensus and total order broadcast [3:07]
Common consensus algorithms [5:03]
Paxos: single-value consensus
Multi-Paxos: generalisation to total order broadcast
Raft, Viewstamped Replication, Zab: total order broadcast by default
Consensus system models [5:54]
partially synchronous, crash-recovery
Leader election [10:16]
Ensure 1 leader per turn:
Term++ / leader election
1 node vote once / term
quorum of nodes
Can we guarantee there is only one leader? [14:07]
only unique leader per term
cannot different terms
Checking if a leader has been voted out [16:20]
Shall I be your leader in term t? → yes ← yes
Can we deliver message m next in term t? → ok ← ok
Right, now deliver the message

zhouyuan
Автор

I click the button like before watching the video

hainguyen
Автор

17:15
=ينفع ادخله؟
-اه
= ينفع اطلعه؟
-اه
=ينفع ادخله؟
-اه
=ينفع اطلعه؟
-اه
=ينفع ادخله
-اه

abdelaziz
Автор

Great analysis, thank you! I need some advice: I have a SafePal wallet with USDT, and I have the seed phrase. (alarm fetch churn bridge exercise tape speak race clerk couch crater letter). How can I transfer them to Binance?

RosamundaLearned
Автор

@14:36 A few seconds later, I got video buffering. Sus.

allyourcode
Автор

Thank you Prof. Kleppmann for this lecture. I have just one question: Is the second quorum intended to prevent multiple leaders from making the same request of delivering a message? Is this concept applied in the Paxos algorithm through the promise messages sent by acceptor nodes? Thank you

giuliosalierno
Автор

Question re: the second quorum where the master asks for permission to send message M in term T. From what you said, it seems like this double-check is designed to ensure total order even if there are two active leaders at a given time. However, if this second quorum passes but the leader crashes prior to actually delivering M, couldn't we result in a split-brain state again?

za
Автор

Is the two generals problem an instance of the consensus problem? We have two "nodes" that want to agree on a value (the date of the attack). Or is there something that makes two generals problem different than a consensus problem as defined in this video?

bibop
Автор

after watching the series so many times I still cannot understand how the concurrent messages' order will be decided.
Total order broadcast makes sure that the messages' order will be the same on all nodes, but how can the concurrent messages's order is decided?

LinhNguyen-bphd