0

I have a home cloud server whose data directory contains about 400GB of data.

I've tried to do a full duplicity backup on the directory. It's been running for 17 hours and only about 19 GB of data (compared to the 368 GB of the previous backup) have been generated. That's a speed of about 310 KB / s (!)

Server :

  • cheap tower from a few years ago
  • Debian 9
  • duplicity from packets (0.7.11)
  • Intel J4205 @1.5Ghz, 1MB cache
  • cheap hard drive but still capable of 50+ MB/s

Backup destination :

  • Synology disk station, also capable of 50+ MB/s (much more as a matter of fact)
  • connected via ethernet (in fact, I can test a file copy independently which runs generally around 100MB/s)
  • A backup passphrase is used, but encryption parameters are kept to default

The CPU seems stuck at 100% on a single duplicity process.

Is this expected? I can understand that my hardware is cheap and a bit old, but is it supposed to be SO slow? I would ideally like a speedup of at least 40-100x by upgrading, how can I even be sure that I can obtain this kind of performance improvement, and what should I even aim for?

PS: it seems that duplicity uses gnuPG which by default uses SHA-256. Still, if it's so slow on a 2016 CPU, it means that you can't even browse the web on this PC?

maxbc
  • 109
  • why the downvote? this question is very precise, provides all the necessary details and may interest anyone who might have questions about duplicity backup performance... – maxbc Jun 04 '21 at 14:59

1 Answers1

0

The CPU isn't very fast (it's low-power), but it's actually not that bad.

May I ask if you want to keep that backup program anyway, or might you consider a different one, with greater features and vastly faster speeds? (PS I cannot make comments, my reputation is too low).

In this case (backup a home Debian server to a NAS-connected LAN) there are a lots of possibilities.

Franco Corbelli
  • 81
  • 3
  • Everything's up for change, but I would like to have minimal setup costs and use mostly mainstream ciphers / archive formats in order to retrieve backups easily, even if the software itself doesn't exist anymore. – maxbc Jul 01 '21 at 15:34
  • In fact I suggest you to use an opensource software (it will exists "forever") with proven backward compatibility of at least 10 years, and that you can find (for example on FreeBSD) directly in the official ports. You can try, maybe you feel good, maybe not. https://github.com/fcorbelli/zpaqfranz or https://sourceforge.net/projects/zpaqfranz/ It has everything you ask for, and much more – Franco Corbelli Jul 02 '21 at 16:18