You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@iotdb.apache.org by Ishani Gupta <is...@gmail.com> on 2026/03/25 13:22:58 UTC
GSoC 2026: Trino-IoTDB Plugin Proposal Architecture [Ishani Gupta]
Hello IoTDB Community,
My name is Ishani Gupta, a CS undergrad at Dronacharya College of
Engineering. I am working on a GSoC 2026 proposal for "Implement
Trino-IoTDB Plugin to enable OLAP on time-series data" and wanted to reach
out early for guidance.
My background is directly relevant: I previously built and merged a Trino +
Apache Iceberg integration at Datazip (PR #511:
https://github.com/datazip-inc/olake/pull/511), where I handled schema
mapping, OLAP query federation, and catalog design on top of object storage.
For the proposal, I am planning the implementation around four core Trino
SPI components:
1. ConnectorMetadata — mapping IoTDB 2.X Table Mode (devices → tables,
measurements → columns)
2. ConnectorSplitManager — partitioning by time range and storage group
3. ConnectorPageSourceProvider — data reading via iotdb-jdbc
4. Predicate pushdown — time-range and device-level filters
Before I finalize the architecture, I have one specific question:
In IoTDB 2.X Table Mode, is the recommended query interface for external
connectors the JDBC driver, the Thrift RPC API, or the REST endpoint? I
want to make sure the connector uses the right layer and does not introduce
unnecessary overhead.
I will share a full proposal draft on this list in the coming days. Any
early feedback from the community or mentors would be very welcome.
Thank you for your time.
Best regards,
Ishani Gupta
ishanigupta1002@gmail.com
https://github.com/Ishani-10
Re: GSoC 2026: Trino-IoTDB Plugin Proposal Architecture [Ishani Gupta]
Posted by Yuan Tian <ja...@gmail.com>.
Hi Ishani,
Thank you for your interest in the Trino-IoTDB Plugin proposal for GSoC
2026.
Regarding your question about the query interface, the JDBC driver is the
preferred choice for external connectors in IoTDB 2.X Table Mode. Using the
JDBC driver will ensure the best compatibility and performance for the
Trino SPI components you've outlined.
I look forward to seeing your full proposal draft soon.
Best regards,
Yuan Tian
On Wed, Mar 25, 2026 at 9:32 PM Ishani Gupta <is...@gmail.com>
wrote:
> Hello IoTDB Community,
>
> My name is Ishani Gupta, a CS undergrad at Dronacharya College of
> Engineering. I am working on a GSoC 2026 proposal for "Implement
> Trino-IoTDB Plugin to enable OLAP on time-series data" and wanted to reach
> out early for guidance.
>
> My background is directly relevant: I previously built and merged a Trino +
> Apache Iceberg integration at Datazip (PR #511:
> https://github.com/datazip-inc/olake/pull/511), where I handled schema
> mapping, OLAP query federation, and catalog design on top of object
> storage.
>
> For the proposal, I am planning the implementation around four core Trino
> SPI components:
> 1. ConnectorMetadata — mapping IoTDB 2.X Table Mode (devices → tables,
> measurements → columns)
> 2. ConnectorSplitManager — partitioning by time range and storage group
> 3. ConnectorPageSourceProvider — data reading via iotdb-jdbc
> 4. Predicate pushdown — time-range and device-level filters
>
> Before I finalize the architecture, I have one specific question:
> In IoTDB 2.X Table Mode, is the recommended query interface for external
> connectors the JDBC driver, the Thrift RPC API, or the REST endpoint? I
> want to make sure the connector uses the right layer and does not introduce
> unnecessary overhead.
>
> I will share a full proposal draft on this list in the coming days. Any
> early feedback from the community or mentors would be very welcome.
>
> Thank you for your time.
>
> Best regards,
> Ishani Gupta
> ishanigupta1002@gmail.com
> https://github.com/Ishani-10
>