上級分類: Distributed data storage

Eventual consistency strategies

Eventual consistency is database state that eventually looks the same everywhere. It assumes a multiple replica database. Each replica is available during a partition. So updates can occur to the same data. You need some way of resynchronizing changed that interfere with eachother

YAML 問題

I have some thoughts and ideas on how to resolve partitions after resynchronization with lost peers.

Essentially we want to merge data that is on every replica together

CRDTs help a bit but do not provide an out of the box solution to conflicting updates. We don't want one person's changes to be lost.

I propose that GUIs of editor interfaces present conflicts with a record simultaneously for resolution.

Which value gets used is in dispute as the data is conflicting.

You kind of need to skip conflicting records! As you don't know which value is correct.

You could have a SLA where you pick the latest update, the majority update or the update that was seen the least and so on.


沒有子分類。

投票 (可選) (別通知) (可選)
請,登錄
語言