Large image imports 18 - 25megs via python script into DICOM instance is very slow

Tested on local services with Windows 8 , 16 Gig with images averaging 18-25megs loads successfully in 1 -2 seconds.

However, running similar DICOM instance on Google CE 16 Gigs in performing remote uploads from command prompt using Orthanc upload Python script gives timeout error or run extremely slow. Similarly, the Orthanc Web Upload is also slow for 25mb images.

I have tested on networks with upload speeds of 100mbs and may conclude Google CE host could be the problem.

python 8042 ./dicomimages

Question: Is there any published benchmarks for Orthanc Storage speeds , I need to upload Mammography Images into Orthanc that are generally 18 megs.

There is no security on this instance and persons are welcome to upload test images.


Two questions:

1) Do you have compression enabled? In our testing, we've found that
disabling compression significantly improves performance (at the
expense of storage space, of course).

2) Have you collected performance metrics from the host this is
running on? CPU Usage, IO usage, IO Contention, etc.? Doing so would
help you identify the bottleneck.

Hi Erik -

Compression was on, I will continue to do additional testing including monitoring CPU, IO metrics, etc.

Please note I’m not using Google Storage at this time and all my tests are done using only the Compute Engine.

Question: If files are loaded on a local instance and then copied via Lua scripts , are these files sent to the remote server compressed ?


Hi Erik -

Resource Monitor shows the following utilization:

100% CPU
0 % Network
1% Memory
DIsk IO 0-500 KB/sec

Network Speed 4 Gigs / Sec


Question: If files are loaded on a local instance and then copied via Lua scripts , are these files sent to the remote server compressed ?

I cannot help you with Google CE, but regarding the question above, the answer is “no”. Orthanc only compresses the files for internal use: Whenever a DICOM file must leave Orthanc, it is decompressed on-the-fly.

Please note that Chris Hafey has made 2 requests related to this question:

However, work has not started on this topic yet.


Hi Sebastien -

Thanks for your quick reply.

It would be nice if Orthanc could use that internal representation of the images (compressed format) and then share that with other Orthanc instances via the Lua script.

With regards to the large files above 20megs , the ISP upload speed over HTTP is a factor and we are currently investigating that was well.

Tests on Google CE directly or locally is fine and averages 2-3 seconds for files up to 25megs.

Hi Erik -

I ran some tests over 100MB download and 10MB upload speed network and the results remains the same for very large images over 24MB.Not sure how to proceed.