[
https://issues.apache.org/jira/browse/PHOENIX-6247?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17305827#comment-17305827
]
Lars Hofhansl commented on PHOENIX-6247:
----------------------------------------
Big fan of this one! It always bothered me that Phoenix did not capitalize on
the ability to abstract Table- and Column- names. We're doing column names now,
so table were missing.
I looked at the PR and found it a bit hard to identify the salient changes from
the boilerplate.
[~giskender] could you point out the core change, or describe the change with a
paragraph, or if you have attach a design doc? Thanks.
> Change SYSTEM.CATALOG to allow separation of physical name (Hbase name) from
> logical name (Phoenix name)
> --------------------------------------------------------------------------------------------------------
>
> Key: PHOENIX-6247
> URL: https://issues.apache.org/jira/browse/PHOENIX-6247
> Project: Phoenix
> Issue Type: Improvement
> Reporter: Gokcen Iskender
> Assignee: Gokcen Iskender
> Priority: Major
>
> Currently, the tables in Phoenix have the same name as the underlying Hbase
> table. Separating logical and physical table name, ie. Having a Phoenix table
> point to an Hbase table with a different name have some advantages.
> An example is this: Let's say we want to have a different storage/encoding
> scheme for an index. We can build the new index while the clients use the old
> index and once the index is rebuilt, we can momentarily start pointing to the
> new index table without much downtime or performance implications. For the
> client, they are using the same index with the same name, but the physical
> table is different. Today, in order to change the index like this, we have to
> drop it and re-create which is a downtime for the index and the data table
> full scans are used for queries impacting performance while the index
> creation goes on.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
