FTPeek_UploadFileFromContainer slow?

[David Wikström]

I'm doing some testing of uploads from container fields (externally stored, open storage), with files in the 30-50 MB range, using FTPeek_UploadFileFromContainer with FTPeek 1.67 on a Mac OS X 10.10.5 with FileMaker Pro Advanced 13 using a local FileMaker file, and I find the upload times to be significantly longer than with an FTP client.

I haven't done any detailed benchmarking on exact differences yet, but one file (54 MB) took 4.5 minutes to uploading using FTPeek (not including the time to connect etc, just the upload itself). Using FileZilla, it took 59 seconds (which already is bad enough, but I don't have a good upload speed where I'm testing this, and neither does my client).

Has anyone else experienced this? 

[David Wikström]

Would the use of a custom memorysize file to increase the head space have any positive effect on file upload time?

[evanseeds]

David, thanks for the information! I'm not aware of any specific outstanding issues with FTPeek that would cause slow uploads, but it's likely that FileZilla client is doing some optimization that FTPeek isn't. I've added a ticket to our issue tracker regarding this sort of upload optimization. If you have any more details that could help, such as the type of FTP server you're uploading to, or any benchmarks you may have done already, those would be a big help for us.

Thanks again!

[David Wikström]

I can't promise that I'll get to this very soon, but I'll try to put together a sample file and do some simple benchmarking to see if the relative upload speed between FTPeek and FileZilla etc seems to vary with file size or not.

[David Wikström]

I did some very rough benchmarking just to see if my initial impression stands or not.

I've uploaded files with different sizes using FTPeek_Upload_Container and FileZilla to an FTP site (not the same as the one I first noticed the issue with), and compared the relative upload times. I ran the uploads 5 times for each file size and the upload times are very consistent for both FTPeek and FileZilla.

I also tried doing the upload with and without a 512m custom memorysize for FTPeek, but that did not make any difference.

Average upload times:

File size              FTPeek            FileZilla             Relative
500k                  00:04                00:04                1:1
5m                     00:28                00:09                3.1:1 
50m                   04:35                00:58                4.75:1

I also ran a 500m file once, and the result indicates that the relative duration doesn't keep increasing (it was 4.6:1).

To me, the above numbers mean that it would make sense to look closer at whether there isn't some built-in bottleneck internal to FTPeek that affects "larger" files on upload. 

I think 360Works would be in a better position to test this properly, but I'd be happy to run some further tests myself. The next thing on my list would probably be to test using another FileMaker-plugin - if another plugin doesn't have the same discrepancy, that would be an ever stronger indication that something isn't as optimized as it can be in FTPeek.

 

 

[David Wikström]

I've now run the same tests using CNS' FTPit Pro 2.0.8, and the results seem to confirm that there's some specific issue with FTPeek.

The relative ratio for the three file sizes is pretty much 1:1 compared to FileZilla; sometimes CNS is even slightly faster, and the very worst I got was 1.36:1 for a 50m file - obviously way better than what I get with FTPeek.