How Do I Improve Write Efficiency of GeminiDB Influx Instances?

1. SDK Lifecycle Management

• Create: In one process, you only need to create a single global instance using the InfluxDB client.
• Use: Call the write or query method without close() after each operation.
• Destroy: Call close() only once when the process is shut down to release resources.

2. Submitting Data Points In Batches

• Advantage: Compared with submission of a single data point, batch submission can significantly minimize network overhead and greatly improve the overall throughput.
• Default SDK configuration: The default batch size of influxdb-java in asynchronous submission mode is 1,000 data points/batch.

3. Optimizing Asynchronous Writes

• Enable asynchronous writes: Enable enableBatch and call the write method.
• After enableBatch is enabled, one client can asynchronously write data to GeminiDB using only one thread. If the write load is too heavy for a single client, you can initialize multiple clients to write data.

InfluxDB influxDB = InfluxDBFactory.connect(serverURL, "username", "password", client);
BatchOptions batchOptions = BatchOptions.DEFAULTS.
//Number of points submitted in each batch
    .actions(1000)
    //Length of the write error buffer queue
    .bufferLimit(20000)
    //Asynchronous submission interval
    .flushDuration(100)
    //Handles a write exception.
    .exceptionHandler((points, throwable) -> {});
influxDB.enableBatch(batchOptions);
influxDB.write(point)

4. Optimizing Synchronous Writes

• Enable synchronous writes: Enable disableBatch and call the write method.
• The performance depends on the number of data points submitted each time. You are advised to submit data points in batches each time.
• To improve the throughput, you can maintain a queue outside the SDK. An external program can control the queue to obtain data in batches and call write() to submit the request.

InfluxDB influxDB = InfluxDBFactory.connect(serverURL, "username", "password", client);
influxDB.disableBatch();
influxDB.write(point)