If an SSD is rated at 1DWPD for 5 years and in actual use one only uses .2DWPD, does that mean that such an ssd would nominally last 25 years barring some other catastrophic hw failure not related to the destructive nature of flash program and erase cycles?
All reliability is about chances of failure. So it’s likely to last that long.
Also we could achieve hyperconvergence "the other way" (unlike having a global shared filesystem like Gluster or Ceph) but use fine grained replication (per VM/VM disk). That's really interesting (data locality, tiering, thin pro etc.). Obviously, we'll collaborate to see how to integrate this in our stack 🙂
This is totally a coincidence that this was the next one up in the queue to post here. But talk about perfect timing! This came up at least three times today alone.
