Question

Causes of different runtimes of the same script in FME Desktop?

  • 16 May 2023
  • 2 replies
  • 0 views

Badge +13
  • Contributor
  • 56 replies

I am asking this question on behalf of my colleagues. My colleagues are facing a problem that they are running a .fmw script at different times of the day. Sometimes this script takes 1 hr and another time this same script takes 2-3 hours. Other applications are also quite slow, so it's probably wise to look at connectivity/databases issues and the likelihood that the problem lies on FME Desktop is probably to a lesser extent or nonexistant. However it could still be a combination of different factors that influence the different runtimes in FME Desktop. So I would like to get a rough idea if there are maybe some issues in FME Desktop that might influence the different runtimes. I was thus wondering if others face different runtimes in FME Desktop as well and might have some ideas about the reasons. One colleague opted some ideas that might impact the different runtimes and I was wondering whether you could stand behind those ideas.

  • Query in SQL databases
  • Transformers in FME Desktop 2022 (maybe slower or inconsistant)vs older versions?

This is more like an open question to get some ideas and in case this question doesn't fit this forum, I will delete this question.


2 replies

Userlevel 6
Badge +31

My experience is that workspace performance is mostly influenced by internal memory and I/O. When the source is quickly accessible (SSD in same workstation) and the ram is sufficient (no "Optimizing Memory Usage. Please wait..." messages) it is as good as it gets.

 

Having 32GB ram, when I see "Optimizing Memory Usage. Please wait..." in the log, I stop the run and redesign the workspace. Optimizing memory means swapping data from ram to disk and this is slow. When I see this it is often an indication of an inefficient developed workspace. Your coworkers might have different types of hardware?

 

Access times of databases can be slow, but I doubt this differs in processing times from one machine to the other. Time of the day, load on database / network might be a thing. In general, if you have a database, you will get better performance if you let the database do the work, instead of loading all data to FME and do it there. Think spatial queries and views.

 

FeatrueCaching, is great for developing phase but comes at a cost when doing larger production runs. Switch it off, because it does eat RAM. It is possible to run with FeatureCaching partly off, using bookmarks and closing them. Personally, I do not like the closed and reopened bookmark behavior so I never use this.

 

Different installations, yes newer installations tend to be faster.

Userlevel 4
Badge +25

Older versions of transformers in your workspace can certainly affect the performance, but I would think that it's across all workspace runs, not a few.

 

Is the workspace processing (roughly) the same amount of data/features on every run?

Reply