Serverless monitoring in Saudi Arabia: Lambda, Cloud Run and Azure Functions traced end to end

A typical serverless application is a chain: an API gateway receives a request, a function validates it and writes to a queue, a second function reads the queue and updates a database, and a third sends a notification. Each step may run on a fresh instance. When the customer waits eight seconds, serverless observability is only useful if the whole chain can be read as one request.

AWS Lambda monitoring. The Datadog Lambda Extension collects function logs through the Lambda Telemetry API, without passing through CloudWatch first, and produces enhanced metrics at a granularity of a few seconds. The library adds traces and custom metrics. For Python and Node.js, Datadog documents trace context carried through API Gateway, SQS, SNS, Kinesis, EventBridge, DynamoDB, S3 and Step Functions, so an asynchronous chain still reads as one request. For other runtimes, or where context cannot be carried, Datadog links related traces instead; we check which applies to your functions before promising a single view.

Relateddistributed tracing page

A cold start is the extra time a platform spends starting a new instance before it can run your code. Users only feel it on some requests, which is why averages hide it. On Lambda, Datadog's init_duration metric measures initialisation time during a cold start, and a cold_start tag marks the affected invocations and traces, so slow start-ups can be counted per function. For Python and Node.js, Datadog also documents cold start tracing, which shows what the function loaded while starting.

On a serverless platform a release that adds 300 milliseconds to a function adds cost to every call. For Lambda, Datadog's estimated_cost enhanced metric gives the estimated cost of each invocation in US dollars, so cost can be charted next to duration and traffic. It is Datadog's estimate, not your AWS invoice. For every platform, Datadog Cloud Cost Management can read the cost exports each cloud writes, so actual spend sits beside usage.

The useful views are simple: the ten functions that cost most per invocation, functions whose cost rose after a deployment, and retries or timeouts that are paid for and deliver nothing.

Monitoring data has its own location question. Datadog is SaaS, and as of November 2026 its documented sites are in the US, Germany, Japan, Australia and the UK, none in the Middle East. Function payloads, logs and tags can carry personal data, so Interkey agrees with your security team what is collected and which fields are removed before anything leaves the function. Support comes from Interkey's engineers in Riyadh, in Arabic and English.

Datadog's bill depends on the number of active functions monitored, the invocations you trace, the traces and logs you index and keep, and the products you add; our Datadog pricing guide sets out how each is billed.

The timeline is an estimate, not a promise. Connecting a cloud and seeing function metrics usually takes days; instrumenting functions through the pipeline and agreeing the tags typically takes a few weeks. The drivers are the number of functions and runtimes, how many teams own them, and how quickly changes pass your release and security reviews.