11-20-2014 05:16 AM
I have an EG project with 2 process flows; with the majority of tasks taking place in 1.
The data sets that are being processed are negligible in terms of size; i.e. the largest being 100,000 observations, 30 variables, most being around 100 observations, 15 variable or less
Problem:In the past couple of days as the project grew to include additional processing steps, any action in the project (think in terms of selecting a task, executing a task, opening a result) takes about 3 to 4 seconds; i.e. if you select a task, you can count 21, 22, 23 and on 24 the task is selected/darkened in the GUI.
Obviously this is rather frustrating.
What can I undertake to alleviate or eliminate the problem
11-20-2014 06:09 AM
It is known that this happens normally with access to DBs with SAS/ACCESS, and sometimes also with simple datasets. The problem is not the machine for sure, or the other processes.
You can always play with Options -> Data -> Performance. Not sure if that option is on EG 4.3. The idea is to reduce the number of rows initially shown to the GUI, then performance improves significantly.
11-20-2014 01:41 PM
Do you have the EG project log switched on? If so try clearing it out. I've found that large project logs can slow down EG jobs.
You could also try setting the maximum number of datasets to open in a project to 0, to avoid this overhead.
11-21-2014 03:46 AM
Thanks for that suggestion SASKiwi. I had not been aware of the Project Log. Still, it was turned off (default) so it's safe to rule that one out as a factor.
As for the data sets, there is only on large one read in but quickly filtered to smaller sizes.
From there on sizes are insignificant (see original post) and they are not opened (i.e. the default setting has been changed to not display result after executing a task.
11-25-2014 03:46 PM
Do you get the same behaviour in a new EG session versus one that has been open for a few hours? I've noticed that over time EG uses more and more memory, and can slow down, possibly as a consequence of using the project log. Does starting with a new EG session help?
Also we have noticed that network latency can slow any task in EG including running a task or opening a dataset. I am assuming here that you work with a remote SAS server.
11-25-2014 04:00 PM
The behaviour was observed in sessions that had not been open very long (half an hour). We did some examination via the task manager but could not find anything in terms of memory hogging, out of the ordinary CPU usage or networking activity.
Network latency does seem like a credible cause. Indeed, I'm working with a remote server and, moreover, the bandwidth for a fixed connection is not impressive given the activity of the offices. I have not had this issue when I worked on the other side of the line and it could also account for the erratic occurrence of this phenomenon; in the past few days I have not observed it.
I'm afraid however that there is no practical means of testing the hypothesis as there are a great number of variables in play here such as virtualization and what not.