In order to improve the performance and efficient usage of our MARS system, the server produces a warning whenever some an inefficient access pattern is detected, eg:
| Code Block | ||
|---|---|---|
| ||
mars - INFO - 20201216.185011 - The efficiency of your requests in the last 12 hours is 14% [marsod-core] mars - WARN - 20201216.185011 - Please see https://confluence.ecmwf.int/x/rxRtD in order to improve the efficiency of your requests [marsod-core] |
A typical inefficient request pattern usually involves repeated access to elements of data which are stored in a single the same file on tape using individual separate MARS requests. This pattern forces multiple mounts of a tape the tape (which is a very expensive action operation for a tape-based mass storage system) which that could be avoided if a single combined larger request were to be used.
The efficiency score is the number of unique tape file accessed divided by the number of requests to access these files:
Access 10 files with 10 requests = 100% score
Access 1 file with 10 requests = 10% score
We would We could recommend you to read Guidelines to write efficient MARS requests, but please do contact our ServiceDesk via email (servicedesk@ecmwf.int) or us via our support portal if you need any further advice.