Interesting results and a bit surprising too. Mira is supposed to be able to run on a potato. Isn't it?
That's what I understood from van's earlier post.
MIRA in hybrid mode replays almost as fast as optimized tmpfs, but since machine doesn’t have enough RAM switching from BMIC to RocksDB takes a significant amount of time.
This looks like a case for spinning up a fresh node from scratch rather than trying to migrate from an existing setup with ram limits as mentioned.
Just curious as I may do just that if it makes a difference to greater decentralisation? If it works?
Yes, it will run and shouldn't slow down just because state can no longer fit in RAM.
That switching I was writing about is on a new node, after replaying in memory. It doesn't fit, so it has to use swap, a lot, thus it takes that much time.
Decentralization itself isn't a value, it need to have a purpose. If you run service that makes use of Steem node, now it will be much cheaper for you to have own node.
Cool, I've been wanting to do a new build to self host my websites for a while now. This could be very useful.
Using swap is fine as long as it has decent speed which this setup seems to show.
The value of decentralisation in my expression above was referring to increased security for the blockchain not the setup above.
Thanks for the update.