I have a workbench that creates a single ECW file. It is intended that this process will be run, from server, up to 100 times, daily.

The problem is that if I execute this workbench from desktop, the process completes in sightly under 1 minute. The exact same process from our server implementation takes 16-17 minutes. THis has a huge impact on our productivity.

The resulting file is just 7Mb and intentionally contains large areas of black, where there was no-data.

I have done some analysis and come to the conclusion that it is the writing to ecw that causes the delay. If I replace the ecw writer with a logger, the rest of the process (reading, mosaicking etc) completes equally quickly in both environments.

This does not appear to be a question of server resources, only a small fraction of the available memory & cpu are used. Also, the temp directory is located in a sensible location, I believe. In both instances we are writing the file locally, so this is not a result of network issues.

Any idea what could be causing this disparity? Is there a difference in how the files are written between the two versions? What could we do to improve the performance on server? Version details are below - we are not in a position to upgrade.

Server - FME Server 2014 SP3 - Build 14391 - linux-x64

Desktop - FME(R) 2014 SP3 (20140814 - Build 14391 - WIN64)

Thanks