Note:
This topic has been translated from a Chinese forum by GPT and might contain errors.
Original topic: tiflash 创建副本粒度疑问？
[TiDB Usage Environment] Production Environment / Testing / PoC
[TiDB Version]
[Reproduction Path] What operations were performed when the issue occurred
[Encountered Issue: Issue Phenomenon and Impact]
[Resource Configuration]
[Attachments: Screenshots / Logs / Monitoring]
Currently, TiFlash supports creating replicas for entire databases and for individual tables. Will it support creating columnar replicas for specified columns in the future? For example, if a table has three fields, can a columnar replica be created for just one of those fields?
The idea is good, but I don’t think it will be implemented this way.
For a single row of data, in TiKV it is just a key and a value. If TiFlash needs to cache a column, it means TiFlash needs to parse the value and pick out a specific column.
Originally, TiFlash regions are replicated based on Raft. If this approach is taken, it means that when TiFlash applies data, it needs to add a filter to retain only certain columns according to the rules.
Additionally, TiDB’s execution plan also needs to change, so that it can use TiFlash when involving this column.
This modification should be quite significant.
Field granularity is not good.
It seems that Oracle’s In-memory DB supports this feature.
Could you describe the specific scenarios in which you want this feature? And what goals do you aim to achieve with this feature?
This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.
|
username: ShawnYan
|
username: 
|
username: 
|
username: 
|
username: 
|
username: