I have been struggeling with OneDiff for some time now.
I am backing up 4 VM`s, 3 server 2012R2 and one Ubuntu.
Ubuntu and two of the 2012R2 servers are a breeze to copy. They are fairly small (40Gb), but our main server, also a 2012R2 is 120Gb with a second disk og 400Gb.
The smaller ones takes about 4-5 minutes each to mirror with OneDiff, while the large VM takes roughly 15 hours to mirror. it takes about 10 hours just to remove snapshots, and the 00001.vmdk files that are created by onediff are almost 100Gb. I have enabled less secure ciphers, but it appears to make no difference in speed. Transfer speed is about 4,5 - 5Mb/s. The system on both sides are esxi 6,7, so it wont accept blowfish ciphers.
So my main question is why the delta files are so huge, even when running onediff subsequently as soon as the first run is done? There should only be small changes to the VM during this time sequence.
The original VM is cluttered with logfiles and other xsi-backup related files from previous backups, both failed and successfull. Are these files in anyway hampering OneDiff?
That has nothing to do with SSH ciphers. By what you describe there's something which is not working quite well, or something you are not taking into account.
As you say, the delta snapshot should not be 100 GB unless you are generating that amount of data in between backups. We don't know if you posted this before activating the cache in your disk controller, in any case that would never explain why the delta snapshot is that big.
Check whether you are adding memory to the snapshot (--snapshot=includememory), which you shouldn't do unless you have some very good reason to. Nonetheless I don't think your memory is 100 GB for that VM.
What you describe seems to be an anomaly on its own. Even if you were not using XSIBackup the snapshot should not be that big.