1

I have set of compressed JPG2K files (https://imap.maryland.gov/Pages/imagery-download.aspx) that I have been converting to IMG. IMG has proven to have 4x faster rendering. The only issue is we have limited HDD space and the IMG aerial files are ~approximately 10x larger.

From what I've read uncompressed files are faster since GeoServer doesn't have to unpack the files. We would like to do some experimenting on compression levels and see if the performance is comparable to IMG.

So for my questions: Would decreasing compression provide any performance increase? Or does the level of compression make no difference to GeoServer?

Improve this question
4
  • 1
    Do it like others with a proven recipe. Read the raster part of slideshare.net/geosolutions/geoserver-on-steroids-foss4g-2015
    – user30184
    Feb 15, 2019 at 15:38
  • I've read through that presentation which is why I switched away from JP2K. The presentation briefly talks about compression but not how the compression level will affect performance. I'm trying to find a middle ground since uncompressed raster performance is much faster in our case but we are running out of HDD space. I have a process for testing layer performance which is why I know IMG is 4x times faster. So I can test compression levels and performance if needed but was hoping to avoid it.
    – mconn
    Feb 15, 2019 at 15:50
  • 1
    Speed depends on the system. Uncompressing takes some CPU time but with compressed images there are less data to read from the file system. Uncompressing JPEG is faster than uncompressing JPEG2000 and there are big differences in the speed of different JPEG2000 libraries. Also with which parameters JPEG2000 compression is done can make a huge difference. You can get some hints by reading the benchmarks from internet but your system may behave differently. From experience I would say that tiled GeoTIFFs compressed with JPEG method to 10-20% of the uncompressed size is often a good compromise.
    – user30184
    Feb 15, 2019 at 16:11
  • Thanks, sounds like I'll have to do some testing. I'll try out those settings and see what the performance difference is.
    – mconn
    Feb 15, 2019 at 16:22

1 Answer 1

Reset to default
3

The usual limiting performance factor is disk access time, rather than compression/uncompression time. So I usually recommend Paul Ramsey's Compression for Dummies approach which creates very small geoTiffs for aerial imagery.

But profiling and testing with JMeter is the only way to be completely certain for your exact case.

Improve this answer
1
  • Thanks, that is a very helpful link. I'll have to do some experimenting.
    – mconn
    Feb 15, 2019 at 21:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.