A small number of customers are reporting that your API gateway is slow. What data can you view to begin troubleshooting the problem?

The best choice for beginning to troubleshoot the performance issue with your API gateway is to review the latency statistics for median, 95th, and 98th percentile of requests in the Cloud Endpoints dashboard. This approach is effective because it allows you to see both general and nuanced performance trends for your API.

Median latency provides a good baseline of typical response times, while the 95th and 98th percentiles give insight into the performance experienced by the slower requests. This data is crucial as it highlights if the slowness is affecting only a small number of outlier requests or if there's a broader issue impacting larger segments of traffic.

Understanding latency metrics can help pinpoint if the problem is due to traffic spikes, processing delays, or perhaps underlying issues in the application or network. By focusing on these percentiles, you can identify potential bottlenecks that might be troubling those specific customers who are reporting slow performance.

In contrast, while reviewing individual request latency or logging every step of the API can provide detailed insights, those methods often require more effort and might be better suited for a deeper dive after identifying general performance trends. Error rates can indicate their own problems, but they do not directly address the slowness issue being reported. Therefore, starting with