Step-by-step: the road to Fleming, 4: Network restarts

This is desirable to insure a homogeneous work load on the vaults. Some machinery must be implemented to tend to this ideal work model, like relocation to section having a bigger XOR space size (or equivalently, a smaller prefix).

I did some simulations with 2 important metrics to optimize about network homogeneity:

  • Number of distinct section XOR space size
  • Density gap (important because a bigger section may not be a problem when it has more nodes)

Yes, when the number of nodes is too big, the section in split in 2 new sections. In current implementation with default min section size (8), this happens when the section has 22 nodes or slightly more. This is not a fixed limit but the probability of a split increases with the number of nodes in the section.

1 Like

Thank you digipl, JPL and tfa for the additional info. The primer did not have much to say about elder relocation, but does nicely address node aging. Sorry, should have read that first. Since the primer does not explicitly exclude them from the aging process, it implies they relocate. The link digipl provided discusses this in more detail. If relocation occurs on a 2^age count of Work Units, nodes (adults and elders) relocate at an exponentially increasing duration as they age, unless they are doing a lot more work as they age. Without a limit, this still seems to give elders too much section stability over time. That stability could be worth building up and selling, like it’s done by gamers. This issue was raised in : Step-by-step: the road to Fleming, 4: Network restarts; in Handling nodes joining; From a section’s point of view. This would be addressed to some extent by section splits, as long as section growth is relatively consistent and elders are split up like adults. Ensuring no node is excluded from being relocated to another section, after some minimum amount of Work Units, and/or time seems worth considering. Maybe I’m still not understanding this process correctly.

BTW, though the SAFE network is taking a long time to go live, those making it happen are doing an impressive job. Resolving the large scale failure problem is a very difficult task. While it could collapse quickly (worst case scenario), it might return slowly to make recovery even more difficult.