Distributed Systems 5.2: Quorums

preview_player
Показать описание
Рекомендации по теме
Комментарии
Автор

absolutely gold, got to understand the whole quorum thing in less than 10 minutes ! thanks Martin!

elvis
Автор

7:18 that diagram is so simple yet helped me understand the r + w > n formula *way* better. 👏

andrewsouthpaw
Автор

Probability of faults
Read-after-write consistency [2:06]
Quorum (2 out of 3) [3:58]
Read and write quorums [5:41]
r + w > n
Read repair [8:19]

zhouyuan
Автор

This is the author of the famous book Designing Data-Intensive Applications (DDIA)!

dwyanewan
Автор

Your are such a legend, thank you so much for this video. I went through my lecture notes continuously, but none of it even remotely came close to how well you explained it. Honestly, couldn't thank you enough and that example was brilliant.

RKelly
Автор

Absolutely pleasure watching it. Nice way of presenting it. Thank you Martin !!

debasishmallick
Автор

Pleasantly clear and simple explanation. Thank you!

Endredos
Автор

very good and well defined explanation. Thanks!

mrmaker
Автор

Beautiful explanation, thank you so much!

omar_m_othman
Автор

Thank you so much for the video, I found it very helpful in understanding the Sloppy Quorum!

Lller
Автор

Vielen Dank fuer die tolle Erklaerung!

welltv
Автор

thanks a lot, Martin for the clear explanation.

sarkersunzidmahmud
Автор

Good job Martin, thanks for shedding light on this topic!

SJM
Автор

at 4:30 what if one of the acknowledgement also fails. We could have another node read the responses and use it, but the initial write doesn't know it's successful, would it write again? it could potentially overwrite the other nodes' work if it does that

abcdef-fotf
Автор

Much appreciated. Thank you! You made this incredibly simple to understand

xxbighotshotxx
Автор

How following problems are solved in replication?

Question 1. Suppose there are 3 nodes.
Variable 'x' has initial value of 0 on all the nodes.
Client triggers operation 'x = 1'.
The operation is successfull on node 1 but fails on nodes 2 and 3 (due to unrelated issue e.g. broken network).
Hence the update operation as a whole has failed but the data on node 1 has changed.
This will lead to inconsist reads - If the quorum includes node 1 then the read(x) will return 1, otherwise read(x) will return 0.


Question 2. In case of multiple variables that form an invariant, one variable might be updated successfully while the other one encounters an issue.

adityadeo
Автор

Question: what if when the quorum write 1 out of 3 succeeded, but 2 failed to write, how can we rollback the one that write ok? Because we already return the client that write if failed.

nealdi
Автор

dude i died, but thnx for explaining i might pass my exams now

ronakjacob
Автор

Are there any real world examples of such data replication exist? It multiplies number of your datastores that you should pay for :c

good-vibes-inc
Автор

Thank you.
- 10 egyptian procrastinators the night before the exam

iyadelwy