[ 
https://issues.apache.org/jira/browse/HBASE-25302?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Xiaolin Ha updated HBASE-25302:
-------------------------------
    Description: 
We have implemented a fast continuous split region method using HFileLink, 
depending on stripe store file manager.

It is very simple and efficiency, we have implement all the ideas and used on 
our production clusters. A region of about 600G can  be splitted to  75G*8 
regions in about five minutes, with less than 50G total rewrite size in the 
whole process, while normal continuous split needs 600G*3=1800G.

Details are in the doc,

[https://docs.google.com/document/d/1hzBMdEFCckw18RE-kQQCe2ArW0MXhmLiiqyqpngItBM/edit?usp=sharing]
    Environment: 
[|https://docs.google.com/document/d/1hzBMdEFCckw18RE-kQQCe2ArW0MXhmLiiqyqpngItBM/edit?usp=sharing]
  (was: We have implemented a fast continuous split region method using 
HFileLink, depending on stripe store file manager.

It is very simple and efficiency, we have implement all the ideas and used on 
our production clusters. A region of about 600G can  be splitted to  75G*8 
regions in about five minutes, with less than 50G total rewrite size in the 
whole process, while normal continuous split needs 600G*3=1800G.

Details are in the doc,

https://docs.google.com/document/d/1hzBMdEFCckw18RE-kQQCe2ArW0MXhmLiiqyqpngItBM/edit?usp=sharing)

> Fast continuous split regions
> -----------------------------
>
>                 Key: HBASE-25302
>                 URL: https://issues.apache.org/jira/browse/HBASE-25302
>             Project: HBase
>          Issue Type: Improvement
>         Environment: 
> [|https://docs.google.com/document/d/1hzBMdEFCckw18RE-kQQCe2ArW0MXhmLiiqyqpngItBM/edit?usp=sharing]
>            Reporter: Xiaolin Ha
>            Assignee: Xiaolin Ha
>            Priority: Major
>
> We have implemented a fast continuous split region method using HFileLink, 
> depending on stripe store file manager.
> It is very simple and efficiency, we have implement all the ideas and used on 
> our production clusters. A region of about 600G can  be splitted to  75G*8 
> regions in about five minutes, with less than 50G total rewrite size in the 
> whole process, while normal continuous split needs 600G*3=1800G.
> Details are in the doc,
> [https://docs.google.com/document/d/1hzBMdEFCckw18RE-kQQCe2ArW0MXhmLiiqyqpngItBM/edit?usp=sharing]



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

Reply via email to