RAM use in SNAP

[insight]

Hello, all
Just upgraded to SNAP 2.0.2, and i m having trouble with RAM use
I tried simple tasks with S1(calibration) datasets and it looks like my 16Gb of RAM get used as expected (almost fully), nevertheless if i perform the same task twice in a row, i noticed it would take much more time the second run.
Only hint i have is when i have a look at the ressource manager, it tells me that my RAM isnt “released” after the first process, before lauching the second process, its almost 80% used.
Had to close SNAP and open again to get memory back .

Any idea on what could cause this ?

NB: I work on a P900 Lenovo worstation with DDR4 ECC RAM

[tolexy]

I’ve been having the same issue. Unfortunately i have to run a process and then quit and restart, quickly subset before moving on with processing. Hopefully someone will have an answer soon…

[marpet]

Actually this should not happen. Even though the memory is used up to 80% or more before the new process is started it should be released as soon as it is needed by the started process.

What you can try is to bring up the performance toolbar.
Select from View / Toolbars / Performance.
In the upper right you will see a graph which shows the memory consumption.
If you click on it the not used memory should be released.

Maybe this helps.

[insight]

Hello, thanks for the answer,
didnt notice that option, thanks for the tip! I will try that

[mohammad.imangholilo]

Hi all,
I would like to share experience so far and hope others have solution to my very specific issue.

My computer is i5 and 4GB RAM. but when doing a process, RAM got to increase to about 95% (normal time about 50 % in use and 1.5-2 GB free) and “not responding” errors comes, also very slow performance.

By searching “Performance parameters” on right top of SNAP desktop window, a guide comes and says about “VM parameters” that:
…these parameters are for advanced users and can be modified only by users having write access on the snap.conf file. …
To change / look at it Tools > Options > Performance in the SNAP.

In short, it seems I don’t have access to write (change the “VM parameters” since the “VM parameters” is not active).
I would kindly suggest you to try this option and maybe you will access to change it.
Would you mind please help me find a solution?

[Nicolas]

If SNAP was installed as an administrator you need administrator right to change these parameters. You can run the “SNAP Configuration optimiser” as administrator or change the etc/snap.conf file from your installation directory.

4GB RAM is very short however to work with large Sentinel products.

[msdpardis]

Hello marpet. I did as you said but it doesn’t release the occupied RAM memory. After splitting two Sentinel 1 images, SNAP doesn’t respond anymore.

[marpet]

If this is the case , I think @lveci should have a look at the split operator.

[Dev]

Hello forum, @marpet @ABraun

I am using SNAP 8.0 and my laptop configuration is 8GB DDR4 RAM, SSD 256GB, 1TB HDD, i5 processor, NVIDIA 4GB GPU

When I process images the RAM usage is going to above 90%

can anybody suggest me how to increase the processing speeds
Should i increase my RAM? If yes, how much?
Is my laptop is capable enough to use SNAP with good processing speed?

Please guide me

Thanks in advance

[marpet]

Hi

yes, 8GB is at the lower end of the requirements. We would recommend 16GB or more.
Here in the FAQ you can read about it.
https://senbox.atlassian.net/wiki/spaces/SNAP/pages/1898053693/SNAP+FAQs#What-are-the-minimal-requirements-for-SNAP%3F

Adding more memory does not necessarily mean a faster processing, but it can help.
Your i5 has probably 4 cores. This means there can be 8 threads in parallel.
With an i7 or so you can have 16 threads at the same time.

But in general you computer is not bad, just a bit short on memory.

In the end more power is in most bases better.
But not always, see my post here:

How to parallelise the snap

In general SNAP does the parallelization automatically, but it does it on a tile basis. The data is split into tiles and each is calculated on a separate thread. But if your data has only 4 or 8 or 2 tiles, then only these number of threads are used. The number of tiles depends on the size of your input images and also how the tiling is defined by the data format. Beside this there are algorithms which prevent a meaningful parallelization based on tiles because they need the full image as input…

[Dev]

Thank you so much @marpet for the reply and information. I appreciate it

[newbee]

Thanks, this is a great tip.

I added the performance toolbar to SNAP and it seems that the max is 650.0MB? Or does that depend on my machine specs?

I’m doing RF classification of a small area with 50 trees and evaluating the feature power set seems to take a long time (over 30 mins).

[marpet]

If this value does not increase, then it seems that you don’t have enough RAM in your computer.
By default, SNAP can use up to 70% of your RAM. Do you have only 1 GB of RAM? For SNAP you should have 8, better 16 GB or even more.
Please have a look at our FAQ

[newbee]

Thanks Marco,

I have 16G of RAM and the max figure seems to fluctuate too

Evaluating the RF feature power set still seems to take a long time but I guess that is expected.

Edit: Just had a look in the snap.conf file in the snap>etc folder and it seems the max for SNAP on my machine is 11G

[marpet]

Maybe the RF does not need more memory. If SNAP needs more memory the max value will increase.
Yes, it takes some time. But I don’t know what an average execution time is. It depends on the data you are using too.