I'm using a random (near the beehive) image from a 383L (5.45um pixels, 3362x2537 pixels 16bit = 17MB) to act as a performance test. The Pentax is a 4" refractor at 670mm native focal length giving a arcsec/pix of 1.66"/pix. The ODroid C2 is a quad core ARM v8 (64bit) with 2GB ram. I'm using a USB2.0 SSD to hold the full set of astrometry.net indices (~32GB).

Astrometry has been setup without in parallel option (the C2 only has 2GB RAM) and you can see from the command line the parameters set - including downsample 4. It's funny that most of the time is in creating the output annotated plots. Successive runs are --overwrite so the solving starts from scratch each time. I was using astrometry with a 4 minutes solve so both runs have the same caching potential from a filing system perspective.

Timed, with annotated plot outputs (top time) = 4 minutes (yikes!)

Timed, without annotated plot outputs (bottom time) ...... 14 seconds (!!!) using one core.

Hi Nick
Sounds like a real good news. It confirm what I thought that Odroid can handle a lot of heavy work in local.
Transmission couldn't be a bottleneck if heavy files can be handled locally.
Just a difficulty to lodge the 32Gb of astrometry files. But is it necessary to have all files availables.

Patrick

You can reduce the number of indices required (hence smaller amount) depending on the depth. However if in doubt!