I have an instance of v.surf.tps running on a host with an
8-core/16-thread CPU. The Gkrellm system monitor shows 0% compound CPU usage
and about 2G of memory being used, but the disk activity is very high. This
suggests a lot of disk reads and writes instead of using more memory (the
host has 32G total).
Are these numbers due to multithreading using all threads, but each one
working off the disk rather than RAM?
I would like to understand so I can speed analyses by maximizing the
hardware capabilities.
Rich
On Mon, 8 Oct 2018, Rich Shepard wrote:
I have an instance of v.surf.tps running on a host with an
8-core/16-thread CPU. The Gkrellm system monitor shows 0% compound CPU
usage and about 2G of memory being used, but the disk activity is very
high. This suggests a lot of disk reads and writes instead of using more
memory (the host has 32G total).
I just killed the process after 125 minutes; most of that time was stuck
on 0% completion at the stage, "Local TPS interpolation with 58 points ..."
on a fast workstation.
The module command line included the input and output file names and the
column name with the precipitation data to be interpolated.
What running time can I expect? How can I find why it's so slow?
Rich
What are the region settings?
···
On 10/08/2018 09:02 PM, Rich Shepard wrote:
On Mon, 8 Oct 2018, Rich Shepard wrote:
I have an instance of v.surf.tps running on a host with an
8-core/16-thread CPU. The Gkrellm system monitor shows 0% compound CPU
usage and about 2G of memory being used, but the disk activity is very
high. This suggests a lot of disk reads and writes instead of using more
memory (the host has 32G total).
I just killed the process after 125 minutes; most of that time was stuck
on 0% completion at the stage, “Local TPS interpolation with 58 points …”
on a fast workstation.
The module command line included the input and output file names and the
column name with the precipitation data to be interpolated.
What running time can I expect? How can I find why it’s so slow?
Rich
grass-user mailing list
grass-user@lists.osgeo.org
https://lists.osgeo.org/mailman/listinfo/grass-user
--
Micha Silver
Ben Gurion Univ.
Sde Boker, Remote Sensing Lab
cell: +972-523-665918
On Mon, 8 Oct 2018, Micha Silver wrote:
What are the region settings?
Micha,
Ah! That explains it: 1.79 billion cells.
The module employed the default 1.5 for thinning, and the default 300M for
memory. I think I'll give it 16000M for memory and let 'er run.
Thanks again,
Rich
On Mon, 8 Oct 2018, Rich Shepard wrote:
The module employed the default 1.5 for thinning, and the default 300M for
memory. I think I'll give it 16000M for memory and let 'er run.
I've learned several things running v.surf.tps on precipitation data from
58 reporting stations. Among them are:
1) Setting cell resolution to 500m, 100m, or 50m makes no difference in
the appearance of the resulting raster map.
2) While the region is set to a boundary, the output extends beyond that
boundary (see attached map).
3) I don't know how to interpret the colors: are blues less than yellows?
All insights into interpreting this output, or modifying the command to
limit output to the region set to the boundary on the map, will help me
better understand how v.surf.tps works.
Rich
(attachments)
