Dear Haonan Hou and the Apache IoTDB Community,

I am writing to formally introduce myself as a candidate for the "Flink
connector for IoTDB 2.X Table Mode" project for GSoC 2026.

As a recent Computer science graduate from CBIT Hyderabad specializing in
IoT and Cybersecurity, I have a strong academic foundation in time-series
data and secure system integration. I am particularly excited about this
project because it addresses a critical bridge in the IoT
ecosystem—enabling real-time stream processing for the modern,
relational-style Table Mode introduced in IoTDB 2.x.

Initial Technical Assessment
After reviewing the iotdb-extras repository and the Flink 1.18+
documentation, I have identified the following key focus areas for my
proposal:
1.) Schema Mapping: Transitioning from the hierarchical Tree Mode to a
structured mapping of Timestamps, Tags (ID columns), and Fields
(Measurement columns) within Flink’s LogicalType system.
2.) Dynamic Table Stack: Implementing the DynamicTableSink and
DynamicTableSource interfaces to ensure the connector is compatible with
Flink SQL and the Table API.
3.) Performance: Utilizing the Tablet interface in the IoTDB Java Session
API to ensure high-throughput batch writes, which is essential for
industrial IoT workloads.
4.) Optimization: Exploring Filter Pushdown (e.g., pushing WHERE clauses
for Tags/Time ranges directly to IoTDB) to minimize data transfer overhead.

Current Progress
I have already begun setting up a local development environment with IoTDB
2.x and Flink 1.18 to test basic connectivity. My next step is to create a
small Proof-of-Concept (PoC) demonstrating a manual mapping of a Flink
RowData object to an IoTDB Table Tablet.

Request for Guidance
As I refine my detailed 350-hour implementation timeline, are there any
specific performance benchmarks or existing architectural constraints
within the 2.x Table Mode that I should prioritize in the early stages of
my proposal?

I look forward to contributing to the Apache Software Foundation and
working closely with the community.

Best regards,
Mahesh Vanekar (mail: [email protected])
https://github.com/Maheshv1204

https://www.linkedin.com/in/mahesh-vanekar-18260226b/

Reply via email to