Temporary Node Supression (as a means to improve transfer time)

There are 7 nodes in my setup, I’m the folder-master for almost all of it.

My machine and one other are very fast in comparison to the rest.

Is there a (non-manual) way to suppress all the other nodes except mine & the other fast machine until these two are sync’d ?

Then remove the suppression and let the remaining nodes benefit from these 2 sync’d sources.

Sorry if this has been discussed before (I did search for it!)

are all nodes connected to each other or just star-like to the folder-master?

They were initially star-like, though now they are all introducers.

That said I’m the folder-master of all but one folder which is small with usually only txt files.

There shouldn’t be a speed advantage to delaying the slower devices further? The fast device should get in sync quickly and then offers the files to the slower devices too.

Currently my max upload speed is 900 KiB/s which is being split between the remaining 6 nodes.

The next fastest machine (the one I don’t want to suppress) has a max upload of 850 KiB/s.

So my thinking was that by syncing to that machine first, I would then be providing the combined upload speed 900 + 850 KiB/s to the remaining (temporarily supressed) five nodes.

I’m hoping this would provide a noticeable (if modest) boost for the remaining 5 nodes.

Is such a boost unlikely / have I a poor understanding of how syncthing works?

Syncthing uses P2P so you will have the best performance/speed if all devices share all the folders with each other. Current limitation: this works for completed files only - but this will be changed in the future - see https://github.com/syncthing/syncthing/issues/950

Thanks for the link, the completed-file-only limit was what I was hoping would benefit from the temporary node suppression.

Rather than spend bandwidth on everyone at once, spend it on providing one other completed copy of the files for one node - and then go back to the normal P2P method for every node.

Ah, I see. I was thinking fast in terms of CPU power and so on. Yes, in your scenario it would be better to first let the other fast device sync and then the others. There’s no mechanism to accomplish that currently, unfortunately.

If the changes are comprised of many files then the default random pull order would make things efficient anyhow. However for single large files this is not the case.

Cheers for the reply!

I’ll manage very happily without anyway :smile:

And now I better check what file pull order I have on all the nodes, I played with those settings a while back & I’m not sure I undid all the changes.

Oh, and thanks to all for the symlink feature, total time-saver!

The new ‘Pause’ feature in v12 is just what I was looking for, thanks a million Syncthing team!

1 Like