Hubic and rclone vs rsync performace

I made two backups of a local 3GB directory of about 17500 files:

  • I rsynced the directory to my ovh web hosting server.
  • I used rclone to copy the same directory to my ovh hubic cloud service.

After modifying two small files I updated my backups.

  • rsync took a couple of minutes
  • ”rclone -u copy” took four hours.

I have the following questions, which I of course should present to hubic but knowing they never answer I present you my questions:

  • is the difference of the performance due to the differences in ordinary file systems and cloud services or because of poor performance of the hubic cloud service?
  • can I improve the performance with any rclone settings?
  • Is there any performance difference in rclone copy/sync?

PS: About hubic

  • I have got no response to my queries.
  • I managed to backup 240 GB of photos and videos with almost the maximum speed of my internet connection (256 kB/s).
  • I use rclone because the hubic linux application sucks: No proper documentation. Crashes. No maintenance. Incomprehensible error messages. No support. Etc.
  • I still try to use hubic, because I paid for a year of 10TB. Which was not necessarily a good idea, but I had so much good experience on other ovh services.

My experience with hubic is that it can be slow at times. Creating containers seems very slow! For instance the rclone integration tests with hubic don’t complete at the moment.

Things to try to make it faster

  • increase --transfers and --checkers - try doubling them until you get lots of errors.
  • grab the latest beta and use the --fast-list flag (assumes you’ve got enough memory for the whole sync listing to fit in).

rclone copy and sync are pretty much the same code so I wouldn’t expect any difference.

hubic is very very slow. rclone is limited to 10 Mbps (in the TOS) but for a single thread I barely get above 240 kB/s. I need --transfers 16 to get 10 Mbps. However, I can run multiple transfers to get multiple 10 Mbps stream but transferring large files is a pain since each thread is so slow and prone to failure

It’s definitely not my connection since it’s from OVH’s datacenters: a dedi in BHS and vps at GRA. From traceroute results it looks like hubic might be in RBX (would love to have some test vps or dedi in RBX :smiley: but can’t find any).