High‑Performance Vehicle IoT Big Data Platform Solution Based on DolphinDB

The vehicle‑IoT ecosystem generates massive GPS trajectory data that, combined with intelligent sensors, IoT, GIS, AI, and OT/IT integration, enables functions such as assisted driving, status monitoring, remote management, and enterprise‑level services like route planning and order tracking.

Traditional time‑series databases can store and query simple series but cannot directly produce the final business results or join time‑series data with business tables, leading to increased system complexity, reduced performance, and higher development difficulty.

A high‑performance vehicle‑IoT big‑data platform must provide four key capabilities: (1) powerful data ingestion supporting multiple protocols and out‑of‑order writes; (2) multi‑modal storage for vehicle status, location, behavior, and sensor data; (3) high‑throughput, low‑latency streaming computation; and (4) strong data analysis tools with visualization.

In a real‑world case, a leading new‑energy vehicle company required handling 1.8 × 10⁸ data points per second and 7 000 measurement points per vehicle. Using DolphinDB, the system achieved 1.8 × 10⁸ writes per second with CPU and memory usage around 40 %, single‑point queries under 100 ms, batch import of historical data at 2.2 × 10⁸ points per second, and support for over 1 600 built‑in functions, while also offering edge‑computing capabilities.

The proposed architecture ingests raw data (time, license plate, latitude, longitude, speed, etc.) into DolphinDB streaming tables, subscribes to these tables, joins with order and vehicle configuration data, and outputs results to application layers, message middleware, or GIS visualization tools.

Deployment steps for a Windows environment include downloading the latest DolphinDB community edition, extracting the zip (no installation required), starting the server, generating 8.64 billion simulated records, and executing performance queries—all within roughly 10 minutes.

Two script files are provided. The first ( // data.txt – database creation and data generation script login(`admin,`123456) // create databases and tables for cars, orders, and drives if(existsDatabase("dfs://t_car")) {dropDatabase("dfs://t_car")} create database "dfs://t_car" partitioned by VALUE([`code]) create table "dfs://t_car"."car" ( code SYMBOL, model SYMBOL, emissions SYMBOL, brand SYMBOL) // ... (omitted for brevity) ... submitJob(`write_data,`write_data,write_data) // background data write ) creates the schema and simulates data, while the second ( // query.txt – performance testing script // load tables and run timed queries timer t=select count(*) from drives // ... (omitted for brevity) ... def orderQuery(oid){ t=select code,btime,etime from loadTable("dfs://t_order",`order) where orderid=oid carcode=t.code[0] tt=select * from loadTable("dfs://dfs_drive",`drive) where code=carcode,ts between t.btime[0]:t.etime[0] return tt } // execute and time the query timer t=orderQuery(1000006) ) defines data generation and eight benchmark query scenarios.

The dataset consists of three tables: t_car (≈100 k vehicle records), t_order (≈1 M orders), and t_drive (≈8.64 × 10⁸ GPS points). The environment configuration lists hardware specs (Windows 11 laptop, i5‑1145G7, 16 GB RAM, 512 GB SSD) and the DolphinDB 2.00.9 server.

Verification notes explain that timers measure server‑side processing only, performance may vary with hardware, and that flushing caches and merging level files ensures accurate benchmarking.