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/
