I have an operational workflow that downloads daily CAMS data (on Rocky Linux). Normally, the job has been completed within 2.5–3 hours.
However, on some days every month, the download is extremely slow and takes 9–12 hours, which disrupts the workflow.
Server: CAMS (ECMWF)
Network: no issues observed locally
My question:
Is this slowdown caused by CAMS/ECMWF server congestion or could it be related to my local configuration ?
This behavior is very likely related to the CAMS/ECMWF side, not your local configuration. A few points to consider:
Server load / congestion: ECMWF services (including CAMS) experience periodic slowdowns during peak times or when large datasets are being requested globally. Several users have reported the same pattern (downloads slowing significantly on certain days).
Local config check: If your local network shows stable throughput (test with iperf, wget from other sources, or speed tests) and the slowdown happens only when pulling from CAMS, it’s unlikely to be a Rocky Linux or local system issue.
Transfer method: Make sure you are using the recommended clients (e.g. ecmwfapi or CDSAPI for CAMS) which support retries and partial downloads. Using plain wget/curl against congested servers often performs worse.
Parallelization: If the workflow allows, splitting requests into smaller chunks and running them in parallel sometimes mitigates slow periods.
In short: this is almost always ECMWF server congestion rather than your configuration. To confirm, you can log download speeds at different times and compare them with ECMWF status updates.
#data-download #cams #network-performance