Improved Backup/Restore

Worldmill has always provided backup/restore to all users. That's good. However, the solution we provided wasn't as great as it could be. In particular, it was slow and we found this to be annoying. Also, it imposed some technical restrictions (because of how it was implemented) which meant that a mill had to be active, but the Foundry service had to be disabled.

Technology is always advancing and since we first implemented Worldmill, many new features have been added to the base cluster technology (kubernetes) that we provide our service upon. We have used one of those new features, volume snapshots, to provide a new backup/restore mechanism.

Pros

  • Faster, usually < 15 seconds to backup or restore
  • Constant Time, independent of volume size or usage

Cons

  • Cannot download backup

Let's talk about the Pros first. Because both backup and restore work at the virtual disk level, the speed is a massive improvement over the current mechanism. With our current approach we could backup (or restore) 1GB of data in about 3 minutes. This was approximately linear, so a Mill with a full standard volume could take up to 30 mins to backup or restore. The new mechanism would do the same thing in approximately 10 seconds.

The only substantial con is that now it is no longer possible to access your backup archive directly and download it to your own machine. I'm not entirely sure how many people made use of that capability, but I think it is worth noting that it is no longer possible. Of course you can still use cyberduck, winscp, etc... to download Mill data to your own machine and we still create a "leaving" zip file when a subscription ends which can be used to download the final image of your Mill.

UI Changes

As part of this work, we've taken the opportunity to re-organise certain aspects of our user interface (UI). It should be simpler now with more operations directly accessible from the "your mills" page.

Show Comments