This document presents new time-based UUID formats which are suited for use as a database key.

  • Ephera
    link
    fedilink
    arrow-up
    6
    ·
    5 months ago

    For others wondering what’s wrong with UUIDv4:

    UUID versions that are not time ordered, such as UUIDv4, have poor database-index locality. This means that new values created in succession are not close to each other in the index; thus, they require inserts to be performed at random locations. The resulting negative performance effects on the common structures used for this (B-tree and its variants) can be dramatic.

    I guess, this means with these new UUIDs, ideally you only create UUIDs on systems that are hooked up to NTP, though I guess, it won’t really be worse than UUIDv4 either way.

    • barsquid@lemmy.world
      link
      fedilink
      arrow-up
      6
      ·
      5 months ago

      My understanding is that as long as IDs are roughly the same range of the index instead of literally random, it reduces the thrashing about needed for indexing these. It probably doesn’t need to be perfectly exact. They’re talking about B-trees, so these would all be modifying the same smaller branches of the tree instead of going in all over the place.

      • Ephera
        link
        fedilink
        arrow-up
        2
        ·
        5 months ago

        Yeah, that is my understanding, too. Otherwise you’d only want to generate them on the database host, as even with NTP there will be small differences. This would kind of defeat the purpose of UUIDs.

        If you’re saying that even without NTP, just by manually setting the time, things will be fine. I mean, maybe. But I’ve seen it far too many times already that some host shows up with 1970-01-01…

        • barsquid@lemmy.world
          link
          fedilink
          arrow-up
          2
          ·
          5 months ago

          Haha, that’s fair, someone absolutely would manage to write 1970s dates if they aren’t pulling the current time.