You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@iotdb.apache.org by "Jialin Qiao (Jira)" <ji...@apache.org> on 2020/02/26 02:31:00 UTC
[jira] [Commented] (IOTDB-523) A potential query optimization for
querying many measurements of a device
[ https://issues.apache.org/jira/browse/IOTDB-523?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17045089#comment-17045089 ]
Jialin Qiao commented on IOTDB-523:
-----------------------------------
Hi, ChunkMetadata is tightly coupled with our query engine. If you want to directly read the Chunks, a new query engine for RawDataQuery may be needed, like the TsFileSequenceRead example.
Another way instead of skipping ChunkMetadata is reading all ChunkMetadata of a TsFile in one time to reduce seek cost.
> A potential query optimization for querying many measurements of a device
> -------------------------------------------------------------------------
>
> Key: IOTDB-523
> URL: https://issues.apache.org/jira/browse/IOTDB-523
> Project: Apache IoTDB
> Issue Type: Improvement
> Components: Core/Engine
> Reporter: Xiangdong Huang
> Priority: Major
>
> Hi,
> Today when I run a SQL like `select * from root` and enable the debug log on server, I find so many DeviceMetaData and ChunkMetadata are read.
> Considering that in TsFile, a device's multiply chunk data will be stored together, is that possible to have the following optimization?
> If a user query many measurements (e.g., select * from, or select more than 1/2 measurements of a device), maybe there is no need to read the ChunkMeatadata... because if we find a ChunkGroup, if one column is needed to be read, other columns are high probability to be read.
> I am not sure whether my idea is correct. Please leave your opinions.
>
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)