Update 'README.md'

README.md

@@ -2,6 +2,29 @@
 
 The Ubuntu server used for backups
 
-## TODO
+## TODO: pull architecture
 
-- Implement pull instead of push backups
+For security reasons, it's better to have a pull architecture for backups instead of push:
+If one server is compromised, the backup server can also be compromised.
+This is especially nasty in case of a ransomware attack.
+After researching, pull is best done using [sshfs](https://borgbackup.readthedocs.io/en/stable/deployment/pull-backup.html).
+NB: using sshfs with our current setup of hard-coded ssh keys seems badly scalable and annoying.
+A better solution could be to use ssh certificates.
+
+One negative side of pull, is that the backup server must now be aware of each server that has data to be backed up.
+I designed our network to be as distributed as possible, such that it is easy to introduce new machines.
+However, most machines that have data to be backed up will be virtual machines in the near future.
+This data can therefore be saved on separate virtual disks.
+The backup server only has to be aware of the physical servers running the hypervisors and back up these disks.
+Fortunately, the amount of physical servers we own is unlikely to change much.
+
+## TODO: off-site backups
+
+Right now, all backup data is still stored on-site.
+One potential place to store backups off-site is Amazon S3.
+Another options would be HiDrive S3 from Strato.
+
+## TODO: offline backups
+
+Another improvement to our backup system is to incorporate offline backups.
+I should investigate what is a good method to do this.