I am migrating from a KinesisProducerLibrary (KPL) to a Kinesis Async Client but the Wrapper class function to this execution is seeing an increase in latency:

Kinesis Async Client:

@Provides
@Singleton
  fun kinesisAsyncClient(): KinesisAsyncClient {

        return KinesisAsyncClient.builder()
            .region(software.amazon.awssdk.regions.Region.of(getAwsRegion()))
            .build()
   }

Use:

try {

            val eventBytes = JSON.writeValueAsBytes(eventRecord)
            val request = PutRecordRequest.builder()
                .streamName(streamName)
                .partitionKey(eventRecord.entityId)
                .data(SdkBytes.fromByteArray(eventBytes))
                .build()

            kinesisAsyncClient.putRecord(request)
        } catch (e: Exception) { ...

I know KPL is good for higher load but I have seen services do far better with Kinesis sync and async clients with similar traffic as mine.

Yet while I am stress testing this logic, the wrapper function latency is increasing as I increase traffic. While Kinesis publish to receive latency is improving.

Am I missing something obvious? Should I add an async client with maxConcurrency value to improve latency?