[
https://issues.apache.org/jira/browse/TAJO-337?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Hyunsik Choi updated TAJO-337:
------------------------------
Description:
Currently, Tajo uses HDFS as a primary storage. But, as a data warehouse
system, Tajo should easily support various data sources.
For this, I propose a generic storage handler interface that provides common
storage methods as follows:
* splitting input data
* locality
* accessing catalog
* creating a table
* removing a table
The above methods are derived from query proecssing mechanism on data sets
stored in HDFS.
Query, SubQuery, and Repartition which usually deal with Fragment should not
deal with concrete Fragment classes. Instead, they should use the methods of
GenericStorageManager.
was:
Currently, Tajo uses HDFS as a primary storage. But, as a data warehouse
system, Tajo should easily support various data sources.
For this, I propose a generic storage handler interface that provides common
storage methods as follows:
* splitting input data
* locality
* accessing catalog
* creating a table
* removing a table
The above methods are derived from query proecssing mechanism on data sets
stored in HDFS.
Query, SubQuery, and Repartition which usually deal with Fragment should not
deal with concrete Fragment classes. Instead, they should use
> Generic StorageManager to provide common storage methods
> --------------------------------------------------------
>
> Key: TAJO-337
> URL: https://issues.apache.org/jira/browse/TAJO-337
> Project: Tajo
> Issue Type: Improvement
> Components: catalog, storage
> Reporter: Hyunsik Choi
> Assignee: Hyunsik Choi
>
> Currently, Tajo uses HDFS as a primary storage. But, as a data warehouse
> system, Tajo should easily support various data sources.
> For this, I propose a generic storage handler interface that provides common
> storage methods as follows:
> * splitting input data
> * locality
> * accessing catalog
> * creating a table
> * removing a table
> The above methods are derived from query proecssing mechanism on data sets
> stored in HDFS.
> Query, SubQuery, and Repartition which usually deal with Fragment should not
> deal with concrete Fragment classes. Instead, they should use the methods of
> GenericStorageManager.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
