Update extension overhead explanation

README.md

@@ -21,11 +21,11 @@ Follow the [configuration instructions](https://docs.datadoghq.com/serverless/co
 
 The Datadog Lambda Extension introduces a small amount of overhead to your Lambda function's cold starts (that is, the higher init duration), as the Extension needs to initialize. Datadog is continuously optimizing the Lambda extension performance and recommend always using the latest release.
 
-You may notice an increase of your Lambda function's reported duration. This is because the Datadog Lambda Extension needs to flush data back to the Datadog API. Although the time spent by the extension flushing data is reported as part of the duration, it's done *after* AWS returns your function's response back to the client. In other words, the added duration *does not slow down* your Lambda function. See this [AWS blog post](https://aws.amazon.com/blogs/compute/performance-and-functionality-improvements-for-aws-lambda-extensions/) for more technical information. To monitor your function's actual performance and exclude the duration added by the Datadog extension, use the metric `aws.lambda.enhanced.runtime_duration`.
+You may notice an increase of your Lambda function's reported duration (`aws.lambda.duration` or `aws.lambda.enhanced.duration`). This is because the Datadog Lambda Extension needs to flush data back to the Datadog API. Although the time spent by the extension flushing data is reported as part of the duration, it's done *after* AWS returns your function's response back to the client. In other words, the added duration *does not slow down* your Lambda function. See this [AWS blog post](https://aws.amazon.com/blogs/compute/performance-and-functionality-improvements-for-aws-lambda-extensions/) for more technical information. To monitor your function's actual performance and exclude the duration added by the Datadog extension, use the metric `aws.lambda.enhanced.runtime_duration`.
 
-By default, the Extension flushes data back to Datadog at the end of each invocation (for example, cold starts always trigger flushing). This avoids delays of data arrival for sporadic invocations from low-traffic applications, cron jobs, and manual tests. Once the Extension detects a steady and frequent invocation pattern (more than once per minute), it batches data from multiple invocations and flushes periodically at the beginning of the invocation when it's due. This means that *the busier your function is, the lower the average duration overhead per invocation*. In other words, for low-traffic applications, the duration overhead would be noticeable while the associated cost overhead is typically negligible; for high-traffic applications, the duration overhead would be barely noticeable.   
+By default, the Extension flushes data back to Datadog at the end of each invocation (for example, cold starts always trigger flushing). This avoids delays of data arrival for sporadic invocations from low-traffic applications, cron jobs, and manual tests. Once the Extension detects a steady and frequent invocation pattern (more than once per minute), it batches data from multiple invocations and flushes periodically at the beginning of the invocation when it's due. This means that *the busier your function is, the lower the average duration overhead per invocation*. In other words, for low-traffic applications, the duration overhead would be noticeable while the associated cost overhead is typically negligible; for high-traffic applications, the duration overhead would be barely noticeable. To understand the duration overhead that is used by the Datadog extension to flush data, use the metric `aws.lambda.post_runtime_extensions_duration` or `aws.lambda.enhanced.post_runtime_duration`. 
 
-For Lambda functions deployed in a region that is far from the Datadog site, for example, a Lambda function deployed in eu-west-1 reporting data to the US1 Datadog site, can observe a higher duration overhead due to the network latency.
+For Lambda functions deployed in a region that is far from the Datadog site, for example, a Lambda function deployed in eu-west-1 reporting data to the US1 Datadog site, can observe a higher duration (and therefore, cost) overhead due to the network latency. To reduce the overhead, configure the extension to flush data less often, such as every minute `DD_SERVERLESS_FLUSH_STRATEGY=periodically,60000`.
 
 ## Opening Issues