Issue running Graph Builder - TileCache ? 8.0.8update?

Hi everybody,

Snap interface is having a weird behavior :
1/ Very slow Graph Builder while customing settings.
2/ Neverending process stucked at 0 or 1% (please find the xml used as example here under)

I don’t know if there is a link but I figured out that there is a warning when I open SNAP at the bottom right corner:imageThis is happening on the 5 computer I’m using…

Is any one facing this issue after the SNAP’s 8.0.8 update (8.0.5-S1TBX)?

I saw this topic and figured out that the TileCache might be an issue. I guessed that the Reset option (Tools>Option>Performance) was cleaning the cache, but it seems just reset the settings to 1024Mb CacheSize (C:\Users\xxx.snap\var\cache isn’t clean).

I finally got the graph builder faster, but I don’t know why.
I managed to get my graph running well in 16minutes, but only once. It is now stuck at 0% since one hour, the output is growing very slowly (Koctet per Ko…)

The CPU is about 10% used, and RAM 20 to 40%.

I’m really confused… any feedback is welcome!

Complex_to_Ground_Detected_Calibrated_Despeckle_ortho_v2.xml (5.7 KB)

The GUI will allocate Java Heap and/or cache, and not release until the GUI exits, so RAM utilization will may not go down. If you complete your graph with the input image, etc., you can run from the command prompt “gpt yourgraph.xml” which may help. I suspect there is something wrong with your SNAP/S1TBX installation: possibly user vs admin installation, or install or cache directory/file permissions. I don’t know how big your COSMO-SKYMED image will be at 2.4m resolution, but adding “subset” prior to write might allow you to run a small area (using less memory) for testing.

Thank you for the help s0sh0rt,

The CSK is about 1.6Go and I don’t have any memory issue thanks to the 64Go of RAM DDR5. I’ve also the admin permissions so that should not be a problem either.

Indeed GPT is running well. I’m avoiding the SNAP GUI as much as possible, but I made this graph for my colleagues in order to run quickly an orthorectification from complex image. Unfortunately it is not quick at all and I don’t get why. It takes about 5minutes using GPT optimisations.


Okay, the xml I sent here is working good indeed.
it seems that the data I was working with was corrupted, I don’t have any other explanation.

Thank you s0sh0rt for the tips of closing the SNAP interface between different runs, it cleans the RAM and speed up for manually fill the later graph builder.

Remains the minor issue “unexpected Exception” at the bottom right corner, I guess I’m not the only one, but it does not affect the processings.


Related comment: Speckle-filtering should be done in radar geometry before terrain correction.

Yes, thanks for the info

Marcus: You note Speckle-filtering should be done in radar geometry. I do not believe this is possible when processing “multi-temporal speckle filters” since the many dates must be co-registered first (using terrain correction?). Is there a method to co-register multiple dates in radar geometry prior to “multi-temporal speckle filtering”?

Co-registration and terrain correction are two different things. There are several ways to co-register in radar geometry including create stack or colocation.